United States
Environmental Protection
Agency
Office of Air Quality
Planning and Standards
Research Triangle Park NC 27711
EPA-450/4-80-020
October 1980
Air
Guideline For Applying
The Airshed Model
To  Urban Areas

-------
                                 EPA-450/4-80-020
  Guideline For Applying The
Airshed  Model to Urban Areas
         U.S. ENVIRONMENTAL PROTECTION AGENCY
            Office of Air, Noise, and Radiation
         Office of Air Quality Planning and Standards
           Monitoring and Data Analysis Division
         Research Triangle Park, North Carolina 27711

                  October 1980

-------
This document 1s Issued by the Environmental Protection Agency to
report technical Information of Interest to a limited number of readers.
Copies are available as supplies permit from the Library Services Office
(MD-35), U.S. Environmental Protection Agency, Research Triangle Park,
NC  27711; or, for a nominal fee, from the National Technical Information
Service, 5285 Port Royal Road. Springfield, VA  22161.
                                         Publication No. EPA-450/4-80-020

-------
                           ACKNOWLEDGEMENTS

     This guideline was prepared by David E.  Lay land of the Source Receptor
Analysis Branch of the Monitoring and Data Analysis  Division of the Office
of A1r Quality Planning and Standards.  John  Summerhays, also of the Source
Receptor Analysis Branch, made Important contributions  to the guideline.
E. L. Martinez and Dr. Henry S. Cole provided additional assistance.
Guidance and direction were given by Joseph A. Tikvart  and David H. Barrett.
Steven D. Reynolds, Thomas W. Tesche, and Lawrence E. Reid of Systems Appli-
cations, Incorporated provided various materials which  were adapted for
Inclusion 1n the guideline.  Appreciation 1s  extended to Zada Nelson for her
care and patience 1n preparing the manuscript.

-------
                               FOREWORD

     High ozone levels are one of the most pervasive air pollution problems
facing our nation's cities.  While the planning and analysis tools available
to grapple with the problem have been less than Ideal, the costs associated
with the types of control measures necessary to alleviate the problem are
high.
     The need for credible modeling techniques for evaluating the relation-
ship between emission reductions and ozone levels on an urban scale 1s widely
recognized.  A credible technique should possess at least two basic attributes,
First, 1t should account for all significant physical and chemical processes
and second, 1t should be capable of being verified.
     Empirical techniques have been proposed and used for developing State
Implementation Plans called for by the Clean A1r Act.  Unfortunately,
empirical techniques have not fully satisfied the criteria of technical
rigor and verlflability.  Photochemical grid models on the other hand meet
both of these criteria.  However, the benefits of using such models are
achieved at appreciable cost in terms of data requirements.
     The physical and chemical processes which lead to high observed ozone
levels are complex.  The relationship between precursor emissions and ozone
air quality 1s time and space dependent.  Among the major factors affecting
the generation of ozone are:
     r the quantity and the spatial  and temporal distribution of emissions
       of nitrogen oxides and volatile organlcs
                                     *
     - the chemical composition of the volatile organlcs emitted
     - the chemical reactions among ozone, organlcs, and nitrogen oxides
                                   11

-------
     - the spatial  and temporal  nature of the wind field
     - the stability of the  atmosphere and the dynamics of the mixed layer
     - the Intensity and diurnal variation of solar radiation
     - the transport of ozone  and  Its precursors Into the urban area from
       aloft and Immediately upwind
     - the loss of  ozone and precursors by surface uptake
     The Environmental Sciences Research Laboratory  of the  Office  of Research
and Development, U. S. Environmental  Protection  Agency, has supported a multi-
year research effort by Systems Applications,  Inc.  (SAI)  of San  Rafael,
California.  This research has led to the development of the Airshed Model.
Initial model development efforts and applications  to the Los Angeles area are
fully described 1n a 15-volume series of reports by Roth, et al.,  (1971)  and
Reynolds, et al., (1973a).  These reports are  summarized 1n three  papers  by
Reynolds, et al., (1973b, 1974) and Roth, et al., (1974).  Evaluation of  the
model's predictions and Its components at that time showed  that Improved  treat-
ments of some physical and chemical processes  were  necessary.  As  a result,
another series of research efforts was carried out  to Improve the  model.   These
efforts are described 1n a seven-volume series of reports by Jerskey and  Seinfeld
(1976), Jerskey, et al., (1976), Klllus, et al., (1977), Lamb (1976), Lamb, et
al., (1977), L1u, et al., (1976a), and Reynolds, et al., (1976) and in papers by
                                    111

-------
Lamb, Chen, and Seinfeld (1975), L1u and Seinfeld (1975),  L1u,  Whitney,
and Roth (1976), and Reynolds (1977).
     Previous versions of the Airshed Model  have been successfully applied 1n
Los Angeles, Denver, and Sacramento.  These  applications are reported  1n  Reynolds,
                                                           ^
et alJ, (1979), Reynolds, et al., (1978), and Anderson, et al., (1977).   Future
applications are planned for Tulsa, Philadelphia, Washington, D.  C., Baltimore,
New York City, and Boston.
     Although this guideline 1s written in terms of applying the  Airshed
Model, the same general principles apply to  the application of  any photo-
chemical grid model.  Mention of the Airshed Model by name does not consti-
tute formal endorsement of or preference for this model at this time.
     The purpose of this guideline 1s to familiarize air pollution control
agencies and others with the Airshed Model and its potential for  use in  State
Implementation Plan (SIP) development and for evaluating alternative strategies
for the control of photochemical oxidants.  This is not a  how-to  document nor
a formal planning document.  Rather, it 1s one which attempts to  outline  what
taskssare Involved 1n using the Airshed Model.  With this  information  it  1s
hoped that control agencies contemplating a  photochemical  model application
will be able to effectively plan for such a  project.  Planning  considera-
tions.will be unique for each individual application.  For example, the  extent
to which In-house versus contract resources  are used must  be considered  on
a case by case basis.  This document only.describes the kinds of tasks involved
and the overall level of resources required  for an Airshed Model  application.
                                   1v

-------
                          TABLE OF CONTENTS

  I.  Principles of the Airshed Model   	  ....	      1
      A.  Model  Concepts and Mechanics   	  .....      1
      B.  Treatment of Significant Processes 	  .      8
          1.  Advection and Dispersion	      8
          2.  Emissions	      9
          3.  Chemistry	     10

 II.  Overview of the Modeling Process 	     15
      A.  Role of Modeling 1n A1r Quality Planning .  .  	     15
      B.  Selection of the Modeling Region 	     17
      C.  Collection of Aerometric and Emissions  Data	     18
      D.  Model  Verification and Control  Strategy Analysis 	     22

III.  Data Needs   	     24
      A.  Source and Emission Inventories  	     24
          1.  Requirements for Source and Emissions Data  	     24
          2.  Emission Projections   	'	     36
          3.  Emissions Data Handling	     40
      B.  A1r Quality Data	     47
          1.  Requirements for Air Quality Data	     47
          2.  Collection of A1r Quality Data 	     49
      C.  Meteorological Data	     53
          1.  Requirements for Meteorological Data 	     54
          2.  Collection of Meteorological Data  	     56

 IV.  Preparation of Model Input Data	     61
      A.  Day Selection Criteria   	     61
      B.  Data Input Files           	     62
          1.  Preparing the Meteorological Data  	     68
          2.  Preparing the Air Quality Data 	     70
          3.  Preparing the Emissions Data   	     74

  V.  Evaluation of Model Performance  	     75
      A.  Graphical Techniques   	     77
      B.  Statistical Techniques 	     81
      C.  Rever1f1 cation   	     84

 VI.  Model Application for Air Quality Planning 	     92
      A.  Preparation of Future Year Emissions   	     93
      B.  Future Year Model Simulation 	     94
          1.  Baseline Simulation  	     95
          2.  Control Strategy Simulation  .	     96
      C.  Interpretation of Model Results  	    106

VII.  Resource Requirements for an Airshed Model  Study	    Ill
      A.  Personnel  	    Ill
      B.  Computer Facility	    114
      C.  Project Timetable   	  .....    115
      D.  Overall Project Costs  	    118

References	    124

Appendix.  Technical Description of the Airshed Model  	    A-l

-------
              CHAPTER I.   PRINCIPLES  OF THE AIRSHED MODEL

A.   MODEL CONCEPTS AND MECHANICS
                 •
     The Airshed Model 1s a photochemical  dispersion model designed  to
simulate the concentrations of ozone  and  Its  precursors  (nitrogen  oxides
and organlcs*) as they evolve during  a day's  episode.  This  Involves using
a detailed set of meteorological  and  emissions  data 1n order to  simulate  the
emissions of precursors,  atmospheric  dispersion,  and the chemical  reactions
that generate ozone.
     The basis of the mathematics for simulating  concentrations  of ozone
and precursors 1s the conservation of mass.   The  major processes which
cause changes 1n the amount of mass 1n a  given  air parcel  are emissions,
advecflon and dispersion  of mass Into and out of  the  parcel, and chemical
changes 1n the composition of the parcel .  Thus photochemical dispersion
models are based on solving the following equation (restated 1n  terms of
concentrations:
                                         (mass  1n)-(mass out) due
Concentration « Initial concentration +  /to advection  and d1spers1on\
                                        *     parcel  volume         '
                        of emissions) +  ch     in C0ncentrat1on  due
                     parcel volume      to chemical transformation     (1)
In the Airshed Model, the urban area 1s divided Into a three dimensional
grid system for which solutions of Equation (1) are derived Individ-
ually for each grid cell.  This process 1s Illustrated 1n Figure 1-1.
     The first term 1n Equation (1) 1s "concentration."  The purpose
of running the Airshed Model 1s to estimate the concentration of ozone
      The term "organlcs" Includes both simple hydrocarbons and oxygenated
organlcs.  However, the term organlcs and the term hydrocarbons are used
Interchangeably.
                                   1

-------
                  (a) SPECIFICATION OF THE GRID
                                Transport
Transport
B Transport
i
2 Transport
§-
r~
Transport
-* —

=




Chemistry
Elevated Emissions
A Transpo
Chemistry
Elevated Emissions
A Transpo
I
Chemistry
Elevated Emissions
* Transpo
I ,
Chemistry
Surface Emissions
I

rt
MMBBHM
•MBBM
rt

t


Transport

Transport
Transport
Transport

t
                                                    Top of Modeling Region
                                                    Top of Mixed Layer
                                                    Ground Surface
                       Surface Removal

                       1 to 10 kilometers
(b) ATMOSPHERIC PROCESSES TREATED IN A COLUMN OF GRID CELLS.
Figure 1-1.  Schematic illustration of the grid used and atmospheric
processes treated in the airshed model (adapted from Reynolds,
Tesche, and Reid, 1978).

-------
In each grid cell  for a particular day.  This requires solving Equation  (1)
not only for ozone but also for  NO,  N02 and five classes of organic com-
pounds.  These five classes of organlcs roughly correspond to paraffins,
oleflns, aromatlcs, aldehydes  and ethylene.  The nature of the five classes
and the rationale  for making the distinction 1s discussed later  1n this
chapter.  Also discussed later are several other species Indirectly affect-
Ing ozone production which may optionally  be simulated 1n the Airshed  Model.
Thus a simulation  using the Airshed  Model  Involves  calculating the con-
centrations of each major species  1n each  grid cell  for each hour being
simulated.  For example, a typical model run may simulate 14 hours  (5  a.m.
to 7 p.m.) using a grid system of  1200 grid cells  (15 cells east-west,
20 cells north-south, and 4 cells  high) or more.   Consequently,  the Airshed
Model makes a large number of calculations in simulating the formation
of ozone on any given day.
     The next term 1n Equation (1)  1s the  Initial  concentration. The
equation suggests  that the concentration at the  end of a time  step  equals
the concentration  at the beginning of the  time  step plus the  changes  1n con-
centration (due to advectlon, dispersion,  emissions, and chemistry)  that
occur during the time step.  The length of a  time step 1s  usually six
minutes or less.  The user must supply the Initial concentrations for the
first time step of the simulation.  The concentrations calculated at the
end of the first time step are then used as the Initial concentrations
at the second time step.  This process continues until the entire
desired time period has been simulated.
     A typical time for starting simulations 1s 5 a.m.  This 1s  prior to
the morning rush hour and so concentrations are still fairly low.  This
makes the model estimates less sensitive to Initial  concentration measure-
ments and more sensitive to the chemical and physical processes  being
                                    3

-------
 simulated during the day.   Thus,  the  Airshed Model  requires a concentration
 at the beginning of the simulation, I.e.,  Initial conditions, for each species
 (03, NO, N02, five classes of organics  and any optional species) for each
 grid cell.
      The next term 1n Equation (1) represents  the effects of winds and
 turbulent mixing.  From the perspective of an  Individual grid cell, the
 wind portion of this term  1s based on the  difference between the pollutant
 concentration transported  1n from upwind cells and  the pollutant concentration
 being transported out of the cell.  The dispersion  portion of this term
 1s based on the concentration gradients between the cell and each of the
 six neighboring cells (above, below,  and four  sides).  From a broader per-
 spective, 1t 1s also necessary to be  able  to estimate the mass which 1s
 advected and diffused Into the modeling region as a whole.  Thus, an Input
 requirement of the Airshed Model  1s a set  of concentrations at the boundary
 of the modeling region, I.e., boundary  conditions.  More specifically, the
 Airshed Model  requires that estimates of the concentrations be Input for
 all  species (I.e., 0., NO, N0«, five  classes of organics and any optional
 species)  for each hour simulated  along  the sides and at the top of the
 modeling region.
      It 1s  generally desirable to define the modeling region such that the
 upwind  boundary 1s,  1n fact,  upwind of  the city, in order that the boundary
 concentrations  are relatively low.  However, 1n many parts of the United
 States,  substantial  concentrations of ozone and other pollutants are trans-
 ported  Into  a city from long  distances  upwind.  This pollutant transport 1s
 reflected In the  boundary  concentrations used  1n the simulation.  Thus, It 1s
often Important to obtain  an  accurate estimate of boundary concentrations.
                                  4

-------
     The third term on the right side  of Equation  (1)  accounts  for  pollutant
emissions.  This term 1s  quite Important,  since  the  purpose of  using  the
Airshed Model 1s to determine the relationship between emissions  of hydro-
carbons and NO  and the resulting concentrations of  ozone.  The model
              n
requires the emissions of these species  to be specified  for each  grid cell
for each hour that 1s simulated.  Most of the emissions  are added to  the
lowest grid cell In the respective column of cells.  However, the Air-
shed Model also has the capability of  adding emissions from elevated  point
sources to elevated grid  cells.  Thus, the Airshed Model Introduces the
emissions at the height at which they  are actually found.
     The final term 1n Equation (1) accounts for the effects of chemical
transformations.  This term expresses  the changes  1n chemical composition
of the atmosphere that occur as hydrocarbons and NOX react 1n the presence
of sunlight to form ozone.  This 1s one  of the most  complex components of the
Airshed Model.  Fortunately, the chemistry of the  Airshed Model 1s  fully
contained within the model, so the user 1s not  required  to supply any
Information about atmospheric chemistry.
     The Airshed Model also has the option for  simulating the concentration
of carbon monoxide.  Since CO 1s relatively nonreactive, simulation of CO
concentrations 1s useful  for checking the accuracy of the model's treatment
of advectlon and dispersion.
     It should be clear from the above discussion that the Airshed Model
1s substantially different from the Gaussian dispersion models typically
used for Inert pollutants.  First, the Airshed Model does not assume  steady
state conditions, since ozone concentrations are a  function of the complete
meteorological and chemical history of the air mass.  Unlike most  Inert
pollutant model applications, the Airshed Model considers the  chemical
                                   5

-------
 Interactions of emissions  from throughout the urban area.  Also, the
 Airshed Model does  not  use the empirically derived dispersion coefficients
 (I.e., slgmas) Inherent 1n the Gaussian formulation.  Instead, the Airshed
 Model uses eddy d1ffus1v1ty coefficients to estimate dispersion.
     The above discussion  has outlined the concepts by which the major physical
 and chemical processes  are considered in the Airshed Model.  In order to
 discuss the mechanics of the Airshed Model, however, 1t 1s necessary to
 discuss grlddlng.   As mentioned previously, a typical application of the
 Airshed Model might use a  grid system of 15 cells by 20 cells horizontally by
 4 cells vertically. Each  cell has constant horizontal dimensions, say 4
 kilometers by 4 kilometers.  On the other hand, the vertical dimensions of the
 cells vary 1n time  according to changes in the mixing height.  The user
 determines how many cells  will be used below the mixing height.  In this
 way the mixed layer 1s  divided Into several cells of equal dimension.  The
 vertical dimensions of  the cells typically range from 25 to 500 meters.
     An Important option in the Airshed Model 1s to use cells above the mixed
 layer.  These cells become significant when the mixing height rises, thereby
 entraining pollutants from these cells Into the mixed layer.  This process,
 known as fumigation, 1s a  frequent occurrence 1n the late morning hours as
 the mixing height rises rapidly.  Early morning emissions  from elevated point
 sources become trapped  in  these cells and are then mixed to the ground later
 in the day.
     Figures I-2(a)  through (c) show the various possibilities for speci-
 fying the vertical  dimensions of cells.  Figure I-2(a) and I-2(b) show the
options of simulating the  mixed layer with and without also simulating an
elevated Inversion  layer.  Figure I-2(c) shows that a third option of the
Airshed Model 1s to  have the top of the modeling region below the mixing
height.   The purpose of this option 1s to reduce the vertical dimensions of
                                  6

-------
 o
               MIXING HEIGHT: REGION TOP
    t /1 / 11 / 11 /
                                                                        REGION TOP
                                                                        MIXING HEIGHT
                                                   fftffftf iff n i i J
(a) MODELING REGION IS THE
  SAME HEIGHT AS MIXED LAYER
(b) MODELING REGION INCLUDES
  CELLS ABOVE MIXED LAYER
                      to
                                             MIXING HEIGHT
                                    r r^i^ T* t i r , •• *
                                              REGION TOP
                     (c) MODELING REGION DOES NOT
                        INCLUDE THE ENTIRE MIXED LAYER
                   Figure 1-2. Several possible options for vertical gridding.

-------
 the cells and thus Improve the vertical  resolution.   However, this option has
 the disadvantage that the mixing  height  is  not within the modeling region
 making 1t difficult to properly simulate the  effect of a I1d on the vertical
 flow of pollutants.
 B.    TREATMENT OF SIGNIFICANT PROCESSES
      The first part of this chapter provided  a general  discussion of the
 concepts and  mechanics of the Airshed  Model.  As  stated previously, the
 major processes considered 1n the model  are advectlon and dispersion, emis-
 sions, and chemistry.  The following discussion 1s Intended to provide a
 more detailed discussion  of each  of these processes.
      1.   Advectlon and Dispersion
      Since the Airshed Model  simulates concentrations 1n fixed grid cells,
 advectlon and dispersion  are simulated by transferring appropriate amounts
 of  pollutants from one cell  to another.   Pollutants are transferred horizontally
 from one cell  to the next as  a function  of  the concentration and the speed of the
 wind.  The flux of material  due to  these factors  1s called advectlon.  Pollutants
 are  transferred between two cells also as a function  of the difference between
 the  concentrations  1n the two cells  and  an  eddy d1ffus1v1ty coefficient.  The
 flux  of  material  due to these factors  1s  called dispersion.  The eddy d1f-
 fuslvlty coefficient expresses the  rate  of  atmospheric mixing caused by
 turbulent eddies.
     The  Airshed Model  also simulates  vertical advectlve transfer.
Vertical  transfer arises  primarily from  two sources:   (1) balancing the
effect of convergence  and/or  divergence  of  horizontal  winds 1n order to
achieve a proper mass  balance  of  air,  and (2) accounting for changes 1n
cell height as the mixing height  changes.   The first  source of vertical
                                    8

-------
transfer must be considered 1n preparing  the wind data for use  1n the
model.  For example, 1f the horizontal winds are converging  1n  a cell,
the wind field used 1n the model must have a sufficient  vertical component
to prevent an artificial accumulation of  pollutants  1n the cell.  The
second source of vertical transfer 1s an  artificial  one  arising from the
mechanics of the simulation program.  As  the cell height Increases, material
located at a given height may be transferred from one cell to another with-
out actually changing Its relative vertical position.
     The Airshed Model has differences 1n Us  treatments of  vertical and
horizontal dispersion.  Because horizontal dispersion 1s generally less
Important than advectlon, the Airshed Model uses a single horizontal eddy
d1ffus1v1ty coefficient.  However, vertical eddy d1ffus1v1ty coefficients
are Individually calculated for each cell and  each time  period  as a function
of atmospheric stability and surface roughness.
     2.  Emissions
     The treatment of emissions 1s straightforward conceptually:  the  emis-
sions Into a cell are added uniformly throughout the cell.   Thus, the  change
1n concentration due to emissions  simply  equals the  mass added  divided
by the volume of the cell.  Ground-level  emissions are added Into the  lowest
level cell.  Elevated emissions are added Into upper level  cells.   This
requires that certain stack parameters  be Input for  the  major point sources.
The stack parameters are used along with  the wind  speed  to  estimate an
effective plume height for each major  point source.   The emissions  are
then added uniformly throughout the (upper level)  cell which 1s calculated
to receive the plume.

-------
      The Airshed Model  1s  designed to treat emissions of eight pollutants:
 NO,  N02  five  classes  of organlcs and CO.  NO  emissions must therefore be
 divided  Into  emissions  of  NO and N02 and organlcs must be divided Into emis-
 sions of five organlcs  classes.  These classes are discussed 1n Section 3.
      A point  of comparison between the Airshed Model and Gaussian dis-
 persion  models  1s the treatment of plume rise.  Like most Inert pollutant
 dispersion models,  the  Airshed Model uses the Brlggs (1971) formulae to
 estimate plume  rise.  However, unlike most Gaussian models, the Airshed
 Model  1s capable of simulating emissions that rise above the mixed layer.
 Inert pollutant models  generally disregard any emissions 1n plumes that
 rise above the  mixing height.  Although the Airshed Model also disregards
 any  emissions 1n plumes that rise above the top of the modeling region,
 the  modeling  region 1n  the Airshed Model can extend above the mixing
 height.   It 1s  thus possible to simulate plumes that rise Into an Inversion
 layer between the mixing height and the top of the modeling region.
      3.   Chemistry
      As  mentioned previously, the chemical mechanism 1s one of the most
 complex  components  of the  Airshed Model.  The chemistry of ozone production
 has  been  the  subject  of Intensive study for well over a decade, and yet
 there  1s  surprisingly little agreement on the 11st of reactions to use 1n
 simulating ozone photochemistry.  Thus, this discussion will only provide
 a general outline of  the photochemistry of ozone production.  The discus-
 sion will then  outline  the major distinguishing features of the chemical
mechanism used  In the Airshed Model.
     The general process by which tropospherk ozone 1s formed 1s Illus-
trated 1n Figure 1-3.   Figure I-3(a) shows the N02-NO-03 cycle. These
reactions may be written as chemical equations:
                                  10

-------
            N02
                       SUNLIGHT
          (a) THE N02-NO-03 CYCLE
               N02
PHOTOCHEMICAL
 BY-PRODUCTS
ORGANIC COMPOUNDS
                                             FREE
                                           RADICALS
          (b) ORGANIC OXIDATION OF NO TO N02 WITH OZONE BUILDUP
             Figure 1-3.  Photochemical production of oxidants.
                                  11

-------
           a.   N02 + sunlight -»• NO  + 0
            •
           b.   0 + 02 -»• 03
           c.   03 + NO -»• N02 + 02
      Ozone 1s  not emitted 1n any measurable  quantity, and  1n fact, reactions
 (a)  and  (b) represent the only significant source of tropospherlc ozone.
 However,  these reactions are fully reversed  by  reaction  (c).  Thus, NO  by
                                                                      n
 Itself will not cause significant  concentrations of ozone.  The only way
 significant concentrations of ozone will  occur  1s for some other species
 to oxidize NO  to N02 without destroying ozone.  This 1s  exactly what 1s
 shown 1n  Figure I-3(b).  This may  be written as:
           d.   Organic radicals + NO -»• NOg +  miscellaneous  products
 The  net result of reactions (a), (b), and (d),  then, 1s  a  recycling of
 N02/N0, a modification of the organic species,  and the generation of an
 excess Og molecule.   That 1s, organic species permit the ozone formation
 step while bypassing the ozone destruction step of the N02 - NO - 03
 cycle.  The result 1s a buildup of the ozone concentration.  Reactions
 (a), (b), and  (c)  are fast enough  that a  balance or equilibrium 1s still
 maintained between the concentrations of  N09, NO, and 0*,  but this balance
                          f                Z          3
 Involves  a much higher ozone concentration than would occur without organlcs,
      The  above four  reactions provide an  overview of the chemical process
 of ozone  formation.   However, the  number  of  reactions which affect this
 process 1s  far larger.   In  fact, the number  of  reactions Involved 1s too
 large  for all  the  reactions  to be  Included 1n photochemical models.  Thus
the chemical reactions  simulated 1n  these models are a simplified rep-
resentation of the Innumerable reactions  that actually occur.
      It 1s more difficult to represent the various reactions of organlcs
than 1t 1s to  represent  the  reactions  of  NOX.   There 1s  substantial
                                   12

-------
agreement on about 10 to 15 NO  reactions  as  being the most  significant
                              A
reactions for describing the fate of NO...   However,  there  1s relatively
                                       ^%
little agreement as to how best to represent  organic reactions.  This  1s
primarily due to the almost Infinite number of  species and reactions that
occur.  In tracing the entire path of decomposition  of just  one organic
species, 1t 1s possible to 11st over 100 reactions of that species  and
the products of Its decomposition.  Moreover, urban  emissions Include  a
wide variety of organic species.  Therefore,  a  simplified  representation
of all these reactions 1s needed.
     The typical means of representing significant organic reactions 1s
to utilize a small number of reactions for about three to  five categories
of organics.  For example, the Uvermore Regional A1r Quality (LIRAQ)
model developed by the Lawrence Llvermore  Laboratory at  Llvermore,  Cali-
fornia, (NacCracken and Sauter (1975), MacCracken et al.,  (1978))  uses
three categories of organics roughly described  as oleflns, paraffins,  and
aldehydes.  Miscellaneous compounds are Included in  the  most similar of
the three categories.  For each of these organics categories, LIRAQ uses
about 10 reactions.  The reactions and the reaction  rates  are chosen to
represent the chemistry of an "average compound" in  the  respective chemical
category.
     The means of representing organics reactions in the Airshed Model
1s somewhat different from that used 1n LIRAQ.   In  particular, the cate-
gories used In the Airshed Model represent total numbers of bonds of
specific bond types rather than total numbers of molecules of specific
molecular types.  In other words, the Airshed Model  treats  Individual  com-
pounds not as molecular units but rather as carbon-bond units.  Several
different carbon-bond types are recognized:  single bonds,  double bonds,
                                  13

-------
 aromatic  bonds,  and  carbonyl  bonds.   (These can be recognized as the
 Identifying bond types  for paraffins, oleflns, aromatlcs and aldehydes.)
 The double  bonds are further  broken down  Into highly reactive and moderately
 reactive  double  bonds.   The single bonds  actually correspond to the number
 of carbon atoms  which are  only singly bonded.  An aromatic bond 1s actually
 treated as  an  aromatic  ring.   Carbonyl bonds Include not only those
 associated  with  aldehydes, but also those associated with ketones and esters.
 The single  bonds are the least reactive while the highly reactive double
 bonds  are,  as  their  description Implies,  the most reactive; the others are of
 Intermediate reactivity.
     Another species which can be considered 1n the Airshed Model Is CO.
 Carbon monoxide  plays only a  minor role 1n ozone photochemistry and most
 photochemical  models consider it Insignificant enough to Ignore.  Carbon
 monoxide  has more significance as a relatively Inert gas which may be used
 as  a tracer.  By comparing the CO concentrations estimated by the model to
 measurements of  CO concentrations, it 1s  possible to obtain a direct
 assessment  of  the accuracy of the treatments of emissions, advectlon, and
 dispersion  without the  complicating Influence of chemistry.  This can be
 quite  useful for assessing model performance.
     Another feature of the chemical mechanism 1s the option to consider
 temperature effects.  Temperature has a significant effect on the rates
 at which many  reactions  occur,  although 1t 1s unclear what effect 1t has on
 overall ozone  production.   Unfortunately, there 1s little known about the
 temperature  dependence of  the  rates of many Important photochemical
 reactions.   Nevertheless,  the Airshed Model does provide the option to
utilize current  knowledge  about  the temperature variation of the rate
constants used 1n the chemical mechanism.
                                  14

-------
             CHAPTER II.  OVERVIEW OF THE MODELING PROCESS
A.   ROLE OF MODELING IN AIR QUALITY PLANNING
     The Clean A1r Act requires the States to prepare State Implementation
Plans (SIPs) for assuring that ambient air quality standards are met within
a reasonable time frame by Implementing various control  measures on a
progressive schedule.  In the case of oxldants, the control measures thought
necessary Involve considerable expense to the community and may require
significant changes 1n transportation and land use practices.  While there
are many elements Involved 1n achieving public acceptance of and support
for such measures, a fundamental element 1s the air quality planning and
analysis which lead to their adoption.  Decision-makers, legislators, and
the concerned public must perceive the planning process as one 1n which
due consideration 1s given to the technical aspects of the oxldant problem.
Success hinges on credibility which In turn requires that recommendations
for action be based on a firm technical foundation.
     Modeling 1s a tool which enables air quality planners to demonstrate
a cause and effect relationship between specific control measures and
anticipated Improvements 1n air quality.  Photochemical grid models, such
as the Airshed Model, simulate the physical and chemical processes which
lead to observed oxldant levels in urban areas.  Model results are
verified against measured ozone data.  The effects of a given emission
reduction not only on peak ozone levels but also on Its spatial and
temporal distribution and on total population exposure can be determined.
Using such a model, the air quality planning agency will have the
                                   15

-------
 Information and tools at Its disposal  to  prepare  and defend with confidence
 Its plan for achieving the ozone standard and  Improving public health.
      Use of the Airshed Model  1s no small  undertaking.  The model requires
 substantial Input data and lengthy computer runs  to operate.  A successful
 modeling study depends on the active participation of the air pollution
 control  agency, the transportation planning agency, the regional planning
 agency,  and all affected city and county  governments.
      The experience of the Association of Bay  Area Governments (metropolitan
 San Francisco) 1n carrying out a detailed photochemical modeling study has
 been summarized 1n a report entitled "Application of Photochemical Models 1n
 the Development of State Implementation Plans, Volume I:  The Use of Photo-
 chemical  Models 1n Urban 0x1dant Studies," (ABAG, 1979).  Although a different
 model, LIRAQ  was used, the planning considerations, data collection activities,
 and modeling aspects of the San Francisco application lend Insight Into the
 use of photochemical  dispersion models more generally.
      While  the actual  model  application and much  of the data collection may be
 done under  contract,  the study direction  necessarily comes from the local
 agencies.   The local  agencies  are also a  primary  source of Information.  The
 development of a suitable emission Inventory for  a base year along with pro-
 jections  of anticipated growth in future years requires the coordinated
 efforts of  the regional  planning agency,  the pollution control agency, and
 the  transportation planning  agency.  Because highway mobile sources are such
 a major contributor to the oxldant problem 1n  urban areas, the use by the
 transportation  agency  of sophisticated transportation and emissions models
 1s a critical element  1n the study.  Collection of supplementary air quality
and meteorological  data  requires  the Involvement  of the pollution control
agency.   The selection of control measures  to  be  evaluated and the selection

                                  16

-------
of a final strategy requires the participation  of all  these  bodies  of
government.  Interaction of all  affected groups will  go far  1n  assuring
that the modeling study meets the needs of the  community and results  1n  an
acceptable plan which, when Implemented, achieves air quality goals.
B.   SELECTION OF THE MODELING REGION
     Before a photochemical modeling study can  begin, 1t 1s  first necessary
to define the extent of the modeling region.   In doing so, a number of criteria
should be observed.  In simplest terms however, the modeling region must be big
enough to encompass the problem.  Unfortunately, few  urban areas have sufficient
              j
ozone monitoring data to determine the spatial  extent of elevated ozone  levels.
     The modeling region should encompass the urban core and outlying suburban
areas which together comprise the urban metropolitan  area.   It  should also
Include significant satellite cities or towns and any major  exurban Industrial
sources of oxldant precursors, e.g., hydrocarbon or nitrogen oxides.   Allowance
should be made to Include areas where future growth 1s anticipated.  In  general,
the modeling region should be large enough so as to Include  all major nearby
upwind sources under meteorological conditions known  to be conducive to  ele-
vated ozone levels.
     Transport of oxldants and precursors Into the metropolitan area from
adjacent airsheds and formation of oxldants downwind of the metropolitan
area are also Important considerations when determining the extent of the
modeling region.  Transported ozone can significantly affect ozone levels 1n
urban areas.  As discussed 1n Chapter III, 1n order to quantify the Importance
of transport, 1t 1s necessary to locate supplementary ozone monitoring  sites
well upwind of the city.  Similarly, ozone monitoring sites downwind of the
urban area are necessary to determine ozone maxima resulting from ozone
                                   17

-------
within the  urban plume.   In general, the modeling region should encompass
both  the  upwind and downwind monitoring sites.
      While  the modeling region must be large enough to adequately study the
urban oxidant problem, too large a region 1s undesirable.  Emission Inven-
tory  costs  go up as the size of the modeling region Increases.  In addition,
computer  resource requirements for modeling are proportional to the total
number of grid squares 1n the modeling region.  A compromise must be reached
between the size of the modeling region and the size of an Individual grid
square.   Grid squares approximately 5 kilometers on a side are needed.
In general, the larger the grid size, the more localized effects are
smoothed  over and the less useful Information 1s provided by the model.
By orienting the modeling region 1n an upwind-downwind direction, 1t may
be possible to reduce the size of the region while retaining the desired
grid  size.  Figure II-I shows the modeling region being used for an Air-
shed  Model  application 1n metropolitan Philadelphia.

C.    COLLECTION OF AEROMETRIC AND EMISSIONS DATA
      Use  of the Airshed Model 1n an urban area requires a large aero-
                          »
metric and  emissions data base.  Existing data bases are almost universally
Inadequate  1n Important respects, thus necessitating the collection sup-
plementary  emissions, air quality, and meteorological data.  The required
spatial resolution of a modeling data base 1s on the order of 5 kilometer
grid  squares while the required temporal resolution 1s on the order
of 1  hour.  Data collection programs must be designed with these modeling
                                 18

-------

















1
/
q-























y
f
\
1*


—





















I
]
r





—




















^








— •





























—



















/








.
/







i
1








A
/








^,
f

















/

^~'«








-^
















J
^


'V






/
r
./••



/
\






i



/s





^(l
'"W



-
/*

^^


/
X




V
>
4




<
^.





c.
*

4



•;
.'
/








\




f
V^
*>





•\
X
/
i



V
T






r»
A
4
J


<
r


s






J\




^

^
V




r




t
s




N






\






"X.



t





x-






\,



/
{
f.s
•I
1
•*•%<



^


^>






T.







s


V1









\


\





^•*







s,
*>


V



,
S





\

L.
^~





^
i
I
•






/

/



i
^
\




^
I
t














>^

;
*~
\
^M




>v




N<
P



•




\
l
T




')



\
I



*^y


/
r
1*ii



tudj





(





^^





i
i
V-


\
7


iMn



ran




i


\
"X,

f
/






•^^


/



Ma



ta




.^



N
\
^"







\

f




Sti



•M




t.—




v9
A








»
^




rly



•»^




_^M




x










1
1



Aref



••• •




h

W








\
>







x»






















\
^T
v
>




J






                                                   county Unas
Figure 11-1. Example of a Modeling Region and Grid Specification for an Airshed Model
Application (adapted from Engineering-Science, 1979b).
                                        19

-------
 needs 1n mind.  The Airshed Model, as discussed 1n Chapter IV, offers extensive
 data preparation programs which help make the best use of the data collected.
     Collection of source and emissions data should focus on upgrading and
 Improving the existing source/emission Inventories.  While the Airshed Model
 accepts all five criteria pollutants (carbon monoxide, sulfur dioxide, partlc-
 ulates, nitrogen oxides, and hydrocarbons), emphasis should be placed on
 sources of hydrocarbons and nitrogen oxides.  As discussed 1n Chapter III,
 three fundamental source types are recognized:  point sources, area sources,
 and line sources.  Efforts should first be directed at developing a compre-
 hensive annual Inventory of Individual point sources and distributed area sources
 for the base year.  Area sources must also be allocated to the grid network
 selected for the modeling region.  At the same time, a base-year, comprehensive
 traffic Inventory Is needed for line sources.  Vehicle miles associated with
 both network traffic and off-network traffic must be accounted for.
     Source by source Information on the species composition of hydrocarbon
 emissions 1s a critical element of the emissions data collection effort.  This
 Information 1s used to allocate the total hydrocarbon emissions to the five
 carbon-bond categories required by the Airshed Model.  This allocation
 effectively sets the reactive potential of the hydrocarbon emissions 1n
 generating ozone.
     Additional data collection 1s necessary to characterize the temporal
 distribution of emissions.  Both seasonal and diurnal variations should be
accounted for.  Also, Information on anticipated growth on a regional
 scale 1s needed to project the base year emissions to future years.  The
final  goal  1s an hourly emissions Inventory of all sources, for the base
                                  20

-------
year and the projection years that 1s  representative  of  a weekday  during
the "oxldant season."
     Collection of aerometrfc (meteorological  and  air quality)  data  should
            »
focus on supplementing the existing monitoring network.  This  1s generally
done In the form of a comprehensive "summer study."   Such a  study  should
last about three months during the oxldant season  so  that a  variety  of
high-ozone episodes are represented.  A1r quality  and meteorological  data
are gathered on a ground-based network consisting  of  both existing and
supplementary monitoring sites.  Continuous ozone, nitrogen  d1ox1de/n1tr1c
oxide, and nonmethane hydrocarbon monitors are used to characterize  ground-
level air quality.  Meteorological Instruments for measuring wind  speed
and direction, temperature, and solar4  Insolation are  variously collocated
with the air quality monitors.  Rawlnsonde and plbal  releases  are  used for
characterizing upper-level winds, mixing height, and  stability.  In  addition,
aircraft overflights are desirable 1n  some metropolitan  areas  in order to
characterize air quality aloft.  Finally, grab samples are  collected and
analyzed for Individual hydrocarbon species in order  to  further characterize
the ambient reactive mix.
     As discussed 1n Chapter III, the  air quality  data 1s needed both to
specify the Initial and boundary conditions for model application  and to
evaluate and verify the model's performance at estimating ozone levels in
the base year.  The meteorological data 1s necessary  to  prepare mixing
heights and a three-dimensional wind field which,  together  with emissions,
are the major driving forces which control the model  predictions.
                                    21

-------
 D.    MODEL  VERIFICATION  AND  CONTROL  STRATEGY ANALYSIS
      One  of the  key  features of using a  photochemical grid model such as the
 Airshed Model  1s  that model  results  are  verifiable:  model estimates of ambient
 ozone concentrations can be  compared directly with observed concentrations.
 Various statistical  and  graphical  performance measures may then be employed
 to  assess whether the model  Is  performing  satisfactorily.
       Model  verification 1s  conducted for  the base year.  In other words, the
 base  year emission Inventory together with the aerometrlc data collected dur-
 ing the base year summer study  are used  to prepare the model Input data.  A
 total  of  eleven data files must be prepared prior to running the simulation
 program;  several  other nondata  files are also required.  The Airshed Model
 Includes  a  variety of data preparation programs.  However, not all the methods
 and options  offered  may  be sufficient or appropriate for a particular appHca-
                                                                            w
 tlon,  necessitating  modifications  of the existing programs or writing of new ones,
      Several days during the summer  study  period should be selected for model
 verification.  Such  factors  as  ozone levels and data completeness should be
 considered,  among others.  Simulation of more than a single day allows one
 to  evaluate  how the  model performs for a range of conditions under which
 high ozone concentrations occur !  Moreover, satisfactory model performance
                                •I
 may not be achievable on  a particular day  causing one to turn to other days.
 The emissions data base  1s not  normally  developed for a specific calendar
 day, but  rather for  a representative weekday.  However, the meteorological
 and air quality data must be prepared separately for each calendar day for
which a simulation 1s to be  run.
     Model verification requires a high  level of skill and judgment on the
part of the modeler  In addition  to an Intimate familiarity with the data base.
                                  22

-------
Verification 1s an Iterative process whereby model  executions are
followed by an evaluation of the model's performance, a diagnosis of
any problems, and a reanalysls of the model  Input data.  Modifications
are made, the model 1s rerun, and the process 1s continued until satis-
factory performance 1s achieved on one  or more days.
     Once the model and data base have  been  verified, the Airshed Model
1s ready to be applied to the analysis  of control strategy options.  This
step Is the culmination of all previous steps where the benefits of the
photochemical modeling study are realized.  The question of what control
measures should be Imposed on which sources  with what level of stringency
may be answered using the Airshed Model as an analysis tool.
     Control strategies are evaluated for a  critical  test day when the
mesoscale meteorology, pollutant transport,  and pollutant carryover (from
previous days) result 1n high ozone levels.   Control  strategies may be
formulated using a sensitivity approach to determine an overall regional
emission reduction or by a more rigorous approach 1n which candidate stra-
tegies are simulated.  Control strategy evaluation should be conducted for
several test days of different meteorological and air quality data.  The
effectiveness of a particular control strategy may vary depending on the
aerometrlc conditions on a particular test day.  The ultimate goal 1s a
control strategy consisting of specific control measures for specific
sources which, by means of model simulation, 1s shown to attain the ozone
standard on each of the high-ozone test days.
                                    23

-------
                        CHAPTER III.   DATA NEEDS
 A.    SOURCE AND EMISSION INVENTORIES
      The Airshed Model  and  other photochemical grid models attempt to
 simulate the production, decay, and transformation of multiple pollutants
 1n  a  reactive atmosphere.  In order to  portray this reactive mix over time,
 1t  1s necessary to  estimate the total emissions  of photochemlcally reactive
 pollutants  and their spatial  and temporal  distribution.  The principal
 reactive pollutants affecting ozone formation are hydrocarbons and nitrogen
 oxides.   The Airshed Model  requires an  hourly breakdown of the total
 ground-based emissions  and  elevated emissions of these pollutants for
 each  grid square.   In addition, emissions of carbon monoxide are also
 useful during the modeling  study.
      1.   Requirements for Source and  Emissions Data
      As  Indicated In Table  III-l, emission sources are generally broken down
 Into  three  categories:   line  sources, area sources, and point sources.  Line
 sources  consist of  motor vehicle emissions from  streets and highways. In terms
 of  transportation planning, line sources  encompass both link VMT (vehicle
 miles traveled)  associated  with vehicle trips on the  transportation network
 and intrazonal  VMT  associated with trips  off the network.  Line sources are the
 single most  Important source  of precursor emissions in urban areas.  Area
 sources  consist  of  numerous small stationary source emissions and off-highway
mobile source emissions.  Point sources consist  of stationary sources of vary-
 ing size which are  Individually Identified in detail.  While the Airshed
Model  Itself requires only  emission rates, collection and storage of the source
                                   24

-------
                              Table  III-1

                    Source  and Emissions Data Needs
Category

Line Sources
                Data
Network Links
                                    hourly link emissions;* endpolnt
                                    locations; vehicle mix and model year
                                    mix; diurnal pattern of VMT, average
                                    speed, and % hot and cold starts; other
                                    auxiliary data; VOC and NOX species data,

                                 Intrazonals (off-network)

                                    hourly zone emissions;* grid
                                    allocation factors; other data as above.
Area Sources
Point Sources
   annual county-level emissions* and
   activity levels; grid allocation factors;
   temporal apportioning factors, VOC
   and NOX species data.


   plant Identification; point Identification;
   annual point emissions;* locations; stack
   parameters; control efficiencies; process
   Identification and operating rates; temporal
   apportioning factors; VOC and NO  species
   data.                           x
 Emissions of HC, NOX, and CO.
                                    25

-------
 data  Indicated  1n Table  III-l  1s needed to perform the computations necessary
 to project emissions to  future years, both to examine the effect of growth
 and of additional control measures, and to prepare the data for model Input.
      The necessary procedures  for developing the kinds of data shown 1n Table III-l
 have  been discussed elsewhere.  "Procedures for the Preparation of Emission
 Inventories for Volatile Organic Compounds, Volume I" (EPA, 1977b) describes
 the methods used for compiling annual Inventories of VOC emissions from point and
 area  sources.   However,  the methods for Inventorying area sources are Intended
 for relatively  large geographic areas, such as counties.  A companion document,
 Volume II, 1s entitled "Emission Inventory Requirements for Photochemical A1r
 Quality Simulation Models," (Pacific Environmental Services, 1979).  It describes
 requirements and methodologies for developing an Inventory of point, area, and
 line  (I.e., highway mobile) sources for use 1n photochemical grid models, such
 as the Airshed  Model.  For point and area sources, the spatial and temporal
 resolution in Volume II  builds on the basic annual Inventory of Volume I.  How-
 ever,-for highway mobile sources, a completely different procedure 1s recom-
 mended which 1s based on the use of transportation models and network emission
 models.  Volume II also  provides a general description of VOC species data.  How-
 ever, the use of such data for an Airshed Model application 1s described here.
 The following provides an overview of the Inventory process and highlights the
 salient features an Inventory  should possess 1f it 1s to be used for an Airshed
 Model application.
      In general, a metropolitan area planning to undertake a photochemical model-
 1ng study should already have  most of the basic source/emissions Inventory
 Information 1n some form.  However, 1t 1s usually necessary to update the point
and area source Inventories, assemble the various data elements for the line
                                  26

-------
source Inventory, develop grid allocation factors  for area  sources,  and  conduct
original surveys of point or area source categories  not currently accounted for.
A good annual Inventory 1s the necessary starting  point for area  and point
sources while a good transportation Inventory 1s necessary  for Hne  sources.
     Updating of the point and area source Inventories should  focus  on sources
of hydrocarbons and nitrogen oxides.  Table II1-2  lists the principal sources
of these pollutants 1n urban areas.  While the point source Inventory
Identifies Individual, discrete processes, the area  source  category  break-
down used depends upon the relative similarity of  different sources. Dif-
ferent allocation parameters, temporal patterns, organic species  compositions,
or emission factors may dictate subdivision of the categories  listed 1n
Table II1-2.  For example, the different solvents  used for  drycleaning  have
different chemical compositions and different reactivities. When allocating
the solvent emissions to the five carbon-bond categories, a stoddard solvent
1s treated differently than perchoroethylene.  In  addition, consideration
should be given to the control measures to be evaluated. For  example,
1f stage I and stage II service station controls  are to be  analyzed
separately, then it may be desirable to subdivide  service station emissions
Into those associated with loading underground tanks and those associated
with automobile tank filling.
     Once the point, area, and line source Inventories are  assembled, the
emissions must be apportioned to hourly periods.   The hourly emissions
should be representative of a typical weekday during the oxldant season
(usually June through September).  Original surveys  of hourly process rates
for the major point sources of hydrocarbons and nitrogen oxides are desirable.
For smaller point sources, plant operating schedules are sufficient to derive
typical hourly emissions, while for area sources  temporal patterns  for each
                                   27

-------
                               Table III-2

            Sources of Hydrocarbon and Nitrogen Oxide Emissions

PETROLEUM REFINERIES

   Process Drains and Wastewater Separators
   Vacuum Producing Systems
   Process Unit Turnaround
   Boilers and Process Heaters
   Catalytic Cracking
   Chemical Treating
   Coking
   Blending
   A1r Blowing
   Fugitive Sources

STORAGE, TRANSPORTATION, AND MARKETING OF PETROLEUM PRODUCTS
   011 and Gas Production Fields
   Natural Gas and Natural Gasoline Processing Plants
   Pipelines
   Petroleum Product Storage
   Ship and Barge Transfer.of Gasoline and Crude 011
   Bulk Gasoline Terminals
   Gasoline Bulk Plants   d
   Service Station Loading
   Service Station Unloading

ORGANIC CHEMICAL MANUFACTUREf

   Process Streams
   Storage and Handling
   Waste Handling
   Fugitive Sources

INDUSTRIAL PROCESSES

   Paint, Varnish, and Ink Manufacture
   Food Processing
   Pharmaceutical Manufacture
   Rubber Products Manufacture
   Plastic Products Manufacture
   Textile Polymers Manufacture
   Coke Production
   Mineral Products Manufacture
   Nitrate Fertilizer Manufacture
   Nitric Add Manufteture
   Explosives  Manufacture

                                     28

-------
Table III-2 (continued)

INDUSTRIAL SURFACE COATING

   Large Appliances
   Magnet Wire
   Automobiles
   Cans
   Metal Colls
   Paper
   Fabrics
   Metal Furniture
   Wood Furniture
   Flat Wood Products
   Other Metal Products

NONINDUSTRIAL SURFACE COATING

   Architectural Coatings
   Auto Reflnlshlng

OTHER SOLVENT USE

   Degreesing (toluene, trlchloroethylene, and other)
   Dry Cleaning (stoddard, perchloroethylene, and other)
   Graphic Arts
   Adheslves
   Cutback Asphalt Paving
   Miscellaneous Commercial/Consumer Solvent
   Pesticides

STATIONARY SOURCE COMBUSTION

   External Fuel Combustion
      Electric Generation
      Industrial Boilers
      Commercial and Institutional Boilers
      Residential Heating
   Internal Fuel Combustion
      Turbine
      Reciprocating Engine
   Solid Waste Combustion
      On-Site Incineration
      Open Burning

HIGHWAY MOBILE SOURCES

   Link VMT
      Light Duty Autos
      Light Duty Trucks
      Heavy Duty Gasoline Trucks
      Heavy Duty Diesel Trucks
      Motorcycles
   Intrazonal VMT
      Same as above

                                      29

-------
Table II1-2 (continued)

OFF-HIGHWAY MOBILE SOURCES

   Railroads
   Aircraft
   Vessels
   Construction Equipment
   Agricultural Equipment
   Other Vehicles
     aPetroleum Product Storage—Includes  all  storage facilities
except those at service stations and gasoline  bulk plants.


      Bulk Terminals—emissions from loading tank trucks and
rail cars.


     °Bulk Plants—emissions from storage  and  transfer.


      Service Station Loading—filling of  underground storage tanks.


     eServ1ce Station Unloading—automobile tank filling.


      Organic Chemical Manufacture—only generic classes of processes
associated with the manufacture of various organic chemicals are listed.
                                     30

-------
type of activity are needed.   For Hne sources,  hour-by-hour traffic counts
representative of various roadway functional  classes  (freeway  versus arterial,
etc.) and locale (central business district  versus  suburban residential, etc.)
are needed to characterize the diurnal  distribution of VMT.
     Except when stack tests  or materials  balance calculations show otherwise,
standard AP-42 (EPA, 1977a) emission  factors  can usually be used  for esti-
mating emissions for point and area sources.   Occasionally other  references
                                                                             *
may be required for certain source categories.   For line sources, the MOBILE1
computer program (EPA, 1978a&b) should be  used for  generating  mobile source
emission factors.  Most point and area source emission factors are assumed
constant with time.  However, emission factors for  highway mobile sources
vary dlurnally as vehicle operating conditions change over time.   Both
average vehicle speed (as correlated  with  traffic loading) and percent  hot
and cold starts show distinct variations from hour  to hour and have a marked
effect on motor vehicle emissions. Therefore, hourly variations  in emissions
from links and intrazonals are a result of hourly changes  1n  both VMT and
emission factors.
     The Airshed Model requires that  hydrocarbon emissions be apportioned  to
the five carbon-bond categories discussed  in Chapter  I  in  order  to properly
treat the varying reactivity  of different  organic species.   It is therefore
necessary to obtain organic species (VOC)  data for  each  source or source
type.  Publications are available which give typical  organic  profiles  for a
wide variety of point and area source types in terms  of the weight or  volume
percent of the Individual organic compounds present [Bucon, Macko, and  Taback
(1978); GHscom (1978); THjonls and  Arledge (1976)].  Highway mobile  source
exhaust profiles depend on whether the vehicle is catalyst equipped and there-
fore on the type of fuel burned.  Exhaust  profiles  are  obtained from the
     *A revision of MOBILE1 1s expected in July, 1980.
                                     31

-------
 same publications as are profiles for the evaporative portion of gasoline
 vehicle emissions and dlesel vehicle exhaust emissions.
     Despite the utility of these publications, probably the best data, at
 least for point sources, would be obtained from responses to questionnaires
 sent out to individual sources.  Much of the published data 1s for the Los
 Angeles area and may not be applicable to other parts of the United States.
 For example, surface coating operations exhibit large variations depending
 on the solvent or coating used.  In addition, gasoline composition 1s sig-
 nificantly different in California than it is elsewhere.  Profiles for gasoline
 storage and handling, as well as for highway vehicles, should reflect the
 local average gasoline composition.  In the case of highway vehicles, published
 profiles may be adjusted depending on the composition of the gasoline (Engineering-
 Science, 1979).  For sources where no data are available at all, an alternative
 may be to use the urban ambient mix of organic species (as discussed in Section
 B.  A1r Quality Data of this chapter) as a surrogate profile.
     A note of caution regarding hydrocarbon emissions 1s in order.  Most
 hydrocarbon emission factors reported 1n the standard reference AP-42 do not
 account for aldehyde emissions.  A few source types have aldehyde emission
 factors reported separately 1n AP-42, but most do not.  Therefore, 1t 1s
 generally necessary to adjust the emissions of sources which emit aldehydes,
 primarily stationary and mobile source combustion, to obtain total hydro-
 carbon emissions.  This may be done as follows:
     Adjusted HC Emissions = HC Emissions (	—	*)
                                           I00-we1ght % aldehyde
*as formaldehyde (HCHO)
This adjustment 1s made prior to apportioning the emissions to the five
carbon bond categories, which 1s discussed below.
                                    32

-------
     The data needed to develop a  profile  1n  terms of the five carbon bond
categories used by the Airshed Model  1s  (1) the weight  percent and  (2) the
molecular weight of each species present.  If the data  are reported 1n terms
of volume percents, 1t may be readily converted to weight percents.  The
data are then used to  compute first, the  number of moles of each species
present, and second, the number of moles of each bond type associated with
each species.  Once the moles of bond types are computed for all species,
they are summed to obtain the total number of moles of  each bond type.  The
data manipulations Involved are described  further 1n the Appendix,  Technical
Description of the Airshed Model.
     While the technique just described  Is recommended  when detailed species
by species data 1s available, 1t may be  found that some summarization of  the
data by organic class has been undertaken  prior to reporting the data.   If so,
the calculations necessary to obtain the five carbon bond categories will
depend on the classes actually reported.   For example,  1f the  data  have  been
summarized 1n terms of paraffins,  oleflns, aldehydes,  ketones, alcohols,
acetates and aromatics, the equations in Table II1-3 could be  used  to approxi-
mate the carbon-bond categories.
     The Airshed Model also requires that nitrogen oxides be apportioned
to the two species, NO and NOg.  Little  data  are  available  regarding the
relative proportion of nitric oxide and  nitrogen  dioxide  1n emissions  from
combustion sources.  However, some data  for  stationary sources has  been
published (MHHgan, et al., 1979) and data  used  1n  preparing  a photo-
chemical Inventory for Tulsa have  been reported  (Engineering-Science,  1979).
     While the above data Items are needed for each  source  to  derive a
temporally resolved Inventory of the various  pollutants required by the
Airshed Model, further data 1s needed to provide the appropriate level  of
                                    33

-------
       Table III-3.  Possible Equations for Estimating Airshed
                     Model Carbon Bond Categories
Carbon Bond Category                          Equation

                         mass oleflns C3+
                         MW oleflns C3+
(1) OLE   =

i9\ ADO   -              mass aromatlcs
(Z) ARO   ~              MW aromatlcs

t^\ rABR  -              mass aldehydes  .  mass ketones  .  mass  acetates
\si WWD  -              MW aldehydesMW ketonesMW acetates

fA\ Dflo   _              mass paraffins  /MW paraffins  - 2\
l4' PAR   "                   fin
-------
spatial resolution.  As discussed previously,  the Airshed  Model  requires,
1n general, a grldded Inventory.   However, the model  will  accept Individual
point sources.  It 1s not necessary or desirable to treat  all  point  sources
Individually for modeling purposes.  Rather, point sources should be
separated Into "major" sources and "minor" sources.  Only  the  major  point
sources are left as Individual sources.  The minor point sources are assigned
to grids and their emissions are  aggregated.   Two criteria should be observed
when selecting the major point sources.  First, the emissions  should be
released Into an elevated layer,  at least 50 meters or more above the surface.
Second, the quantity of hydrocarbon or nitrogen oxide emissions  should rep-
resent a significant fraction of  the regional  total emissions, on the order
of one half of one percent.  In this regard,  1t 1s not so  much the emissions
from an Individual source, but rather the total amount of  elevated emissions
within a column of grid cells that 1s Important.
     As with minor point sources, line sources and area sources  must be
assigned to the grid squares.  Link emissions  are readily  assigned on the
basis of their endpolnt coordinates.  Where a  link falls in two or more grids,
the emissions may be apportioned  on the basis  of the fraction of the length
of the link In each grid.  Intrazonal line source emissions may be apportioned
according to the fraction of the area of the zone in each  grid.
     A variety of surrogate Indicators are used to apportion county area
source emissions to the grid squares.  For example, population might be
used for apportioning domestic solvent emissions, commercial land use for
gasoline handling, and basic employment (as compared with service employ-
ment) for degreasing emissions.  In this example, population, commercial
land use, and basic employment are termed allocation parameters.  The
fraction of the county total of each parameter associated with each  grid
                                    35

-------
 square 1s  an  allocation  factor.   Obviously,  this  approach presumes that data
 for the allocation  parameters  are available  at  a  "sub-county" level.  Census
 data along with  transportation,  land  use,  and economic studies oftentimes
 provide such  data for Individual  analysis  zones of sufficient spatial resolu-
 tion for use  1n  photochemical  modeling.  Of  course,  1f the specific locations
 of area sources  are known,  the use of surrogates  1s  unnecessary.  This may
                                                                       *
 be the case for  dlesel locomotives, where  track mileage and switchyard
 locations  are known,  aircraft  emissions where airport locations and relative
 levels of  activity  are known,  and gasoline handling  where a survey of service
 stations has  been undertaken.

      2.  Emission Projections
      A good base year emission Inventory as  described above is necessary
 in order to verify  the model's performance (Chapter  V, Evaluation of
 Model  Performance).   However,  Its primary  importance is actually as a
 starting point from which to project  future  emissions.  Emission projections
 serve  as the  basis  for systematic planning to attain air quality goals.  Two
 types  of projected  inventories are recognized:  (1)  baseline projection
 inventories,  and (2)  strategy  projection Inventories.
     Baseline projections Include the effect of growth and control regula-
 tions  now  1n  effect.  Their purpose 1s to  estimate where an urban area will
 be  1n  terms of emissions and air  quality at  some  future date 1f no additional
 controls are  adopted.  The  need for additional  controls can then be assessed.
While no one  has a crystal  ball,  1t 1s possible with suitable techniques
and assumptions  to estimate future emissions 1n a meaningful way.
                                  36

-------
     Baseline projections are usually done separately for point,  area,  and line
sources.  Point and area source projections are generally the  responsibility of
the air pollution control agency while line source  projections should be  done
by the transportation planning agency.  Coordination  1s  necessary to assure
that projections are done on a consistent basis.
     For consistency, all growth projections,  whether point, area, or line,
should be derived from the same basic social and economic Indicators.   Such
Indicators Include population, employment, housing  and land use.   Regional
growth models, such as IPEF (Interactive Population/Employment Forecasting
Model) are used to forecast regional  population and employment.  These
forecasts are then used together with Information on  land use, transportation,
and planning policy to forecast land  use, population, employment, and  housing
1n Individual transportation zones with a model such  as  PLUM  (Projectlve
Land Use Model).  These models are available from the Federal  Highway  Admin-
istration and are often used by transportation planners  as basic Inputs to
the transportation modeling process.   For projections of stationary source
emissions such forecasts are Invaluable, not only 1n  and of themselves, but
also as regional control totals.  Ideally, projections of both mobile and
stationary source emissions, at a regional level, should be firmly tied to
these fundamental Indicators of growth.
     Transportation models use these  social and economic Indicators to
estimate trip generation and distribution 1n future years.  Once these
trips are assigned to a planned future highway network, the resulting
traffic loadings can be used to calculate emissions.   (If the transportation
                                    37

-------
 model  has not been run for precisely the same year(s)  which  Interest air
 quality planners,  1t may be possible to Interpolate from  the results for
 years  which have been run.   If this  1s  attempted  however, 1t should be done
 by the transportation planning agency to assure Its validity.)  Revised
 vehicle speeds appropriate for loadings on  the future  network should be
 used 1n selecting  emission factors,  as  should revised  percent cold starts
 and hot starts in  areas where land use  1s projected to shift substantially.
     Point and area source projections  are  usually  done on a source category
 by source category basis using the best Information available.  Changes 1n
 gasoline marketing might be estimated on the  basis  of  total  projected VMT
 (from  the transportation model) and  average projected  fuel economy.  Changes
 in petroleum refining might be based on a judgmental application of national
 trends, on projected changes 1n local basic employment, or on the basis of a
 local  survey of the Industry.   Unless Information to the  contrary 1s available,
 1t is  generally necessary to assume  that all  emission  points 1n a large
 Industrial  facility "grow"  by the  same  factor.  Only for  a few sources, such
 as  power plants, 1s sufficient Information  generally available on planned new
 sources  or  facility expansions to  make  possible the actual addition of new
 emission  points  to  the Inventory.
     Normally overall  growth  and the effect of control measures on area source
 emissions are projected  at  the county level (or at  a redefined "political
jurisdiction" level) while  changes in area  source spatial  patterns are handled
 through the use of  allocation  factors developed specifically for a certain
projection year.  Most area source categories  are projected  using such
surrogates as population, employment and  land  use forecasts  and are then

                                     38

-------
allocated to grids using the same surrogates.   It 1s  therefore  possible  to
develop corresponding allocation factors for a specific  projection year.
     Strategy projections, as discussed 1n Chapter VI, Model  Application
for A1r Quality Planning, are used to evaluate the effectiveness  of  addi-
tional proposed control measures and to test the final strategy selected
for attainment of the ozone standard.  Typical control measures that may be
considered Include reasonably available control technology (RACT), lowest
achievable emission rate (LAEff), and best available retrofit  technology  (BART)
for point and area sources.  Inspection and maintenance  (I/M) and transporta-
tion control measures (TCMs) may be considered for line  sources.
     The level of control represented by RACT, for example, when  applied to
a given source category, may be expressed as a control  factor which  1s then
used to reduce the projected (baseline) emissions to arrive at  the controlled
(strategy) emissions.  When controls already exist, 1t 1s probably best to
start from the uncontrolled level and then apply the appropriate  reduction.
A complication arises when, for example, 1t 1s desired to apply a level  of
control corresponding to LAER to new sources while applying a level  cor-
responding to RACT to existing sources of the  same category.   If growth 1s
anticipated, yet specific Information on new units does not exist or 1s
unavailable, 1t may be necessary to make some  simplifying assumptions about
how much of the growth 1s "new" and how much 1s due to an Increase within
existing capacity and therefore "old."  When making longer-term projections,
say to 1995, replacement of existing capacity may be another complicating factor,
     While the simplifications and assumptions required are not Insignifi-
cant, the degree of detail with which projections are performed should  be
commensurate with the accuracy of the growth Indicators from which they are
derived.  Moreover, what 1s Important from a modeling viewpoint 1s the  total
regional change 1n emissions and Its spatial and temporal distribution.
                                    39

-------
      3.   Emissions  Data  Handling
      The  above discussion  has outlined the wide variety of data required for
 the  base  year and projection year  Inventories.  This data 1s used to generate
 an hourly Inventory of eight pollutants  (NO, N02,  five carbon-bond categories,
 and  CO) for each grid  square 1n the modeling region along with an Inventory
 of hourly emissions for  each major point source.
      Assume for the moment that the modeling region contains 400 grid squares
 and  50 major point  sources.  For a fourteen hour simulation the Airshed Model
 requires  emissions  for eight pollutants  for each hour.  The number of emis-
 sion values Input to the model 1s  therefore 400 grids x 8 pollutants x 14
 hours/day plus 50 points x 8 pollutants  x 14 hours/day or a total of 50,400
 values.   Each of these 50,400 values  1s  generated  from a series of calcula-
 tions.  These 50,400 values would  then have to be  regenerated for each base-
 line projection and strategy projection  to be simulated.  Obviously, some
 form of computer assisted  procedure 1s necessary to perform the literally
 millions  of calculations and manipulations necessary.
      A schematic representation of a  computerized  procedure for performing
 the  necessary calculations and converting the results to model format 1s
 shown 1n  Figure III-l.   The procedure requires as  Input an EIS/P&R file of
 point and area source  emissions.   Due to Its design, the EIS/P&R system
 (EPA, 1975) 1s quite flexible and  well suited as the starting point from
which an Airshed Inventory of point and  area source emissions may be
developed.  Line sources should be handled separately however; this 1s
discussed later.
                                   40

-------
                   REPORTS
                                                                    REPORTS
                   TEMPORAL
                     AND
                  POLLUTANT
                    SPLIT
        TEMPORAL
        FACTORS
        POLLUTANT
        PROFILES
  EIS/MR
HOURLY SOURCE/
 EMISSIONS FILE

 MAJOR/MINOR
 POINT SOURCE
DIVISION/MODEL
 CONVERSION
                                                          CONTROL
                                                            CONTROL
                                               AREASOURCE
                                              QRIDDINOmOOEL
                                               CONVERSION
                                            AREA SOURCES
                                           TIME INTERVAL/
                                            GRID VALUES
                                                              REPORTS   I
Figure  III-1.   A POINT  AND  AREA SOURCE DATA HANDLING  PROCEDURE.

                                                  41

-------
      The EIS/P&R file has  a  hierarchical  structure.   Each  facility or
 "plant"  may contain  several  "points"  defined  as  stacks  or  boilers.  Each
 point may contain several  processes or  "SCCs."   The  SCC or Source Clas-
 sification Code  (EPA, 1976)  Identifies  the  kind  of process venting to a
 stack or, 1n the case of boilers,  the type  of boiler and the  fuel burned.
 For  point sources, stack data,  operating  schedule Information, control
 efficiency, and  annual  estimated emissions  are stored at the  point level
 while the annual  fuel/process operating rate  and emission  factor are stored
 at the process level.  Area  sources are readily  treated as "pseudo-point
 sources."  In other  words, the  estimated  annual  county  emissions from each
 area source category are stored on the  file at the point level.  Such an
 EIS/P&R  "master  file" 1s required  for the base year  and for each baseline
 projection or strategy projection.
      The manipulations  necessary to generate  the Airshed inventory are the
 same regardless  of whether the  EIS/P&R  master file represents the base year
 file or  one of several  projection  files.  The procedure Itself consists of
 three  programs each  of  which performs several major  functions.  The first
 program  shown 1n  Figure III-l performs  two  functions:   (1) the annual
                                   i
 emissions  are apportioned to 24 consecutive one-hour periods  and (2) the
 hydrocarbon emissions are apportioned to  the  five carbon-bond categories and
 the  nitrogen oxide emissions are split  Into NO and NOg.  Two  Input files are
 required  to perform  these functions:  a temporal factor file  and a pollutant
profile file.  The temporal  factor file contains a seasonal apportioning
factor, a daily apportioning factor,  and  a  set of 24 hourly apportioning
factors for each source.  In the case of  a  point source for which no temporal

                                      42

-------
factors are Input, the operating schedule Information on  the  EIS/P&R master
file 1s used Instead.  The pollutant profiles may be Input  source  by source,
or by source category.  The profiles contain the  weight percent  of NO  and  N02
and weight percents for each of the five carbon-bond categories.   (The develop-
ment of these percents 1s described 1n  the Appendix, Technical Description of
the Airshed Model.)  The pollutant profile file also contains the  weight per-
cent of aldehydes.  This 1s used to perform the aldehyde  correction discussed
1n Section 1.  Requirements for Source  and Emissions Data.
     The second program 1n Figure III-l  divides up the pdlnt  sources Into
major sources and minor sources.  The user specifies a ton  per year cut-off
for both NO  and HC and a plume height  cut-off.   Point sources exceeding
           A
these cut-offs then become major sources.  The  program then takes  the  major
sources and creates two files in Airshed Model  format, one  containing  the
stack data for each source and the other containing the  corresponding  hour-
by-hour emissions.  The minor sources are assigned to the modeling region
grid, the grid being defined by the user, and a file 1s  created  in model
format containing hour-by-hour emissions for each grid  square.
     The third program takes the county area source emissions and dis-
aggregates it Into gridded emissions.  The user Inputs  a file containing
a set of allocation factors for each allocation parameter.   These factors
represent the fraction of each parameter, e.g., population, in each grid
square.  The user also specifies which  allocation parameter 1s to be used
for allocating which area source category, e.g.,  population,  for domestic
solvent use.  The program then creates  a file  in  model  format containing
hour-by-hour area source emissions for each grid  square.
                                 43

-------
      While a  data  handling  procedure  of the  type just described 1s suitable
 for point and area source emissions,  a  different procedure 1s required for
 line sources.  Figure III-2 shows  a conceptual  flow diagram for such a
 procedure.  The procedure consists of several parts.  First, as shown
 across  the top of  the figure,  link emissions are calculated.  Intrazonal
 emissions are also calculated  and  then  combined with the link emissions when
 the emissions are  allocated to grid squares.
      The  computation  of link emissions  begins with a file of transportation
 network data, a Federal  Highway Administration  (FHWA) "historical record"
 file for  example.   Diurnal  traffic distributions are then used to compute
 hourly  link traffic volumes.   These volumes  are used along with link capacity
 to  estimate hour-by-hour link  speeds.   The volumes are also used to compute
 hour-by-hour  VMT.   This  data,  along with the other transportation data, 1s
 stored  on the "Hourly Link  File."  Next, the MOBI LEI program 1s used to compute
 hour-by-hour  emission factors  for  each  link.  Speeds are obtained from the
 "Hourly Link  File"  while percent cold starts and hot starts, ambient tempera-
 ture, and a variety of other Input parameters required by MOBILE1 are Input
 separately.   The percent cold  starts and hot starts are Input for each
 traffic zone,  land  use classification,  or locale type.  A preprocessor
 program selects the necessary  MOBILE! parameters for each link.  The result
 1s an hour-by-hour, I1nk-by-l1nk emission factor file.  By multiplying these
 emission  factors times the  hour-by-hour VMTs from the "Hourly Link File,"
 the hourly link emissions are  computed.  A similar procedure 1s used to
compute Intrazonal  emissions.
     Because evaporative emissions have a distinctly different organic
composition than do exhaust emissions,  evaporative and exhaust emissions
                                   44

-------
CJ1
                BASE OR
               PROJECTION
                 YEAR
                 LINK
                  FILE
CALCULATE
 HOURLY
LINK DATA
                         CALCULATE
                         HOURLY LINK
                          EMISSIONS
HOURLY
 LINK
 FILE
                                   HOURLY
                                   TRAFFIC
                                  DISTRIBUTION
                                                  HOURLY
                                                   LINK
                                                  EMISSION
                                                  FACTOR
                                                   FILE
                                   CONTROL
                                   STRATEGY
                                                            OTHER
                                                         TRAFFIC DATA
  BASE OR
PROJECTION
   YEAR
INTRAZONAL
   FILE
 CALCULATE
HOURLY ZONE
 EMISSIONS
                                                 GRID LINK
                                                 AND ZONE
                                                 EMISSIONS
HOURLY
  ZONE
EMISSIONS
                                                                                                               GRID
                                                                                                          COORDINATES AND
                                                                                                             ALLOCATION
                                                                                                              FACTORS
                CONTROL
                STRATEGY
                                                            HOURLY
                                                             ZONE
                                                            EMISSION
                                                            FACTOR
                                                                           POLLUTANT
                                                                           PROFILES
                OTHER
             TRAFFIC DATA

( LINE SOURCES
TIME INTERVAL
GRID VALUES


CONVERT TO
MODEL '
FORMAT


                                                                                     HOURLY
                                                                                     GRIDDEO
                                                                                     EMISSIONS
                                                                                       FILE
                                                                             APPLY
                                                                           POLLUTANT
                                                                           PROFILES
                                               Figure 111-2.  A Line Source Data Handling Procedure.

-------
are accounted for separately throughout the procedure.  Therefore, the hourly
emission factor files contain both an exhaust and an evaporative emission
factor.  The result, once the link and Intrazontal emissions are merged, 1s
the Gridded Hourly Emissions File, which contains emissions of exhaust HC,
evaporative HC, NO  and CO.
                  rt
     Next, the exhaust HC and evaporative HC are allocated to the five
carbon-bond categories on the basis of user Input data.  Also, NO  1s split
                                                                 A
Into NO and N02.  The resulting file, containing all eight pollutants, 1s
then converted to model format.
     The hypothetical procedure shown 1n Figure II1-2 1s Intended for
Illustrative purposes.  Each  transportation agency 1s likely to select
and/or develop Its own procedure and corresponding software.  Several
existing network emission models, as the above type of procedure 1s
often termed, are discussed 1n "Volume II.  Emission Inventory Require-
ments for Photochemical A1r Quality Simulation Models" (Pacific Environ-
mental Services, 1979).
                                 46

-------
B.   AIR QUALITY DATA
     In order to simulate the time history of pollutant  concentrations,
1t 1s necessary to know or estimate the mass  of pollutants  entering  from
outside the modeling region.   These "boundary conditions,"  expressed as
concentrations, represent the transport of pollution,  both  ozone  and pre-
cursors, Into an urban area from upwind sources.   In addition,  "Initial
conditions" must be specified as discussed 1n Chapter  I, Principles  of the
Airshed Model.  In other words, the mass of pollutants existing within each
grid cell of the modeling region at the start of the simulation must be  known
or estimated, again 1n terms  of concentration.  Lastly,  verification and
evaluation of the model's air quality estimates requires a  suitable  data
base of measured air quality  for comparison.   Because  existing  monitoring
networks are generally not designed to provide boundary  conditions,  Initial
conditions or adequate verification data, supplementary  monitoring 1s usually
required.
     1.  Requirements for Air Quality Data
         As Indicated 1n Table III-4, air quality data are needed to establish
boundary conditions and Initial conditions for model  application  and to
verify the model results.  The pollutants of  primary concern are  ozone and
Its precursors:  nitrogen dioxide, nitric oxide, and nonmethane hydrocarbons.
In addition, the composition  of nonmethane hydrocarbons  1n terms  of Individ-
ual organic species 1s needed.  Data for other pollutants, I.e.,  carbon
monoxide, are of secondary Importance only.
     The monitoring data base should characterize the following aspects of the
photochemical oxldant problem:
                                47

-------
                                Table III-4

                           A1r Quality Data Needs
Boundary Conditions
Ground-level concentrations on
upwind boundary of region, con-
centrations aloft, and temporal
variation of 03, N02, NO and NMHC.
Initial Conditions
Early morning ground-level con-
centrations throughout region,
concentrations aloft, and vertical
profiles of 03, N02, NO, and NMHC.
Verification Data
Hourly ground-level concentrations
of Oo downwind and within urban
area.
Miscellaneous Data
Species composition of NMHC aloft
and on the ground upwind and within
urban area; concentrations of CO.
at existing monitors.
 Used to enhance both boundary conditions and Initial condition data.
                                  48

-------
     1.  Ozone and precursor concentrations  transported  Into the urban area
from upwind source regions.
     2.  Precursor concentrations within  the urban area  as a result of
local emissions.
     3.  Ozone concentrations within  the  urban area.
     4.  Maximum ozone concentrations downwind of the  urban area.
The first two aspects need to be characterized 1n order  to specify the
Initial and boundary conditions  for Input to the model.  The third and fourth
provide a basis for evaluating the  model's estimates of  ozone  concentrations.
Ozone concentrations downwind are of concern because this  1s where maximum
ozone levels generally occur. Ozone levels  within the urban area are also
of concern because this 1s where population  exposures  are  typically  the
greatest.  In addition, precursor concentrations within  the  urban area are
useful for evaluating the model's estimates  of hydrocarbon and nitrogen
oxide concentrations.
     2.  Collection of A1r Quality  Data
         The data collection program should  be designed  on the basis of
the upwind, urban, and downwind  considerations just  described.  The  siting
of additional monitors 1s almost always required.   Before  supplemental
monitoring stations are Installed,  however,  existing stations  and Instru-
mentation should be used to best advantage.   Relocation  of existing sta-
tions should be considered 1f this  will Improve  the monitoring network.
Typically the data collection effort lasts three months, chosen to coincide
with the "oxldant season."  Historical ozone/oxldant data  should be con-
sulted, although the highest ozone  levels generally occur between June and
September 1n most regions of the United States.
     The number and location of upwind (and downwind) monitoring stations
depends on the consistency of the wind flow from any particular sector
                                  49

-------
 during high ozone periods.   Historical  records of air quality and meteor-
 ology should be examined to Identify any  prevailing wind directions on days
 with high measured ozone concentrations.  The time of day should be con-
 sidered 1n this regard since early morning  surface winds tend to be light and
 variable whereas mid-day winds  are more uniform and better represent the
 regional air flow pattern.   If  a  single wind direction quadrant 1s Identified,
 one or two upwind monitoring stations will  probably suffice; otherwise more may
 be required.  The monitors  should be located fifteen to thirty kilometers
 upwind of the edge of the metropolitan  area and be sited so as to avoid the
 effects of any localized emission sources,  Including highways.  Ozone, nitric
 oxide, nitrogen dioxide, and nonmethane hydrocarbons should be measured at
 each upwind station.
      Urban monitoring stations  should be  located so as to characterize the
 spatial distribution  of ozone and precursors 1n the metropolitan area.
 The number of stations depends  on the area's size and the configuration of Its
 emission sources.   At least two stations  should be located in the central
 business district.  Other sites should  be selected so as to achieve a
 balanced representation of  the  emissions  1n the metropolitan area and to
 monitor areas of special  Interest (urban  residential, suburban, or Industrial
 areas, for example).   The total number  of urban sites may range from as few
 as  three to as  many as nine.  Again,  ozone, nitric oxide, nitrogen dioxide,
 and  nonmethane  hydrocarbons should be measured at each urban site.
     Additional  ozone monitors  should be  located downwind 1n order to
 determine  ozone maxima under conditions of  downwind transport.  Their
 number and location depends,  as for upwind  monitors, on the existence of
 a predominant wind direction  during periods of high ozone concentrations.
 Ozone.maxima  frequently occur fifteen to  thirty kilometers or more down-
wind of  the urban core.  The  total  number of downwind stations may range
                                    50

-------
from as few as four where the wind direction  1s highly persistent to as
many as nine or more where 1t 1s not.   At  least one station should be
located beyond the expected area of maximum ozone  Impact to assure that
the network actually encompasses the highest  values.  Downwind stations
should be sited away from areas of significant emissions, Including Isolated
point sources and highways.  If ozone  levels  1n downwind suburban towns  or
cities are desired to be measured, additional stations would be  needed.
     If nitrogen dioxide 1s now or 1s  expected to  be  a problem,  additional
Instrumentation at the downwind stations nearest the  urban area  should  be
considered.  Some evidence Indicates that  maximum  nitrogen dioxide  levels
may occur five to fifteen kilometers downwind of the  urban core.
     In addition to the basic monitoring network,  Individual organic  species
should be measured 1n order to determine the  ambient  mix of carbon-bonds.
Generally grab samples are taken over  a three hour period and then  analyzed
              * •                                      &
1n the laboratory by gas chromatography for species concentrations. J"he
species data totals also serve as a check  on  the nonmethane hydrocarbon
measurements.  Grab samples should be  taken at the urban monitoring stations
                                                          /*
and 1f transported hydrocarbons are expected  to be significant,  at the up-
wind stations as well.
     Figure III-3 shows an air quality monitoring  network  for a  major
metropolitan area having a well-defined predominant wind direction from the
west and southwest.  In this case, the number of  upwind,  urban,  and downwind
stations 1s 2, 7, and 7, respectively. Ozone 1s measured  at all 16 sites
while NOX and NMHC are measured both upwind and within the urban area.
Due to the general urbanization of the area as  a whole, NMHC 1s also measured
                                   51

-------
         PA
                                                     \
                                ScW     •03,NOX         \

                                r                       _)
                                ,3. NO,    BBS"**-*-
                                ---               °3-N
-------
at two near downwind locations  while  NO and NOg 1s recorded both near and
far downwind.  In this example, several other pollutants are measured
which are not always Included 1n a  photochemical monitoring program.
These are HN03, PAN, and CO.  The nature of the monitoring network for any
particular city will depend on  local  factors, Including the size of the
urban area, the location of existing  monitors, the Importance of Inter-
urban transport, and the prevailing wind dlrectlon(s)  during periods of
high ozone concentrations.
     Development of a monitoring data base Imposes quality assurance
requirements more stringent than for  routine monitoring purposes.  Site
exposure should be 1n conformance with established guidelines  (Ludwlg  and
Shelar, 1978).  Routine Instrument  checks, Including zero and  span,  should
be performed twice per week during  the three-month study  period and  Inde-
pendent audits should be performed  once  a  month.  These quality assurance
measures should be conducted  at both  the existing and supplemental monitoring
stations.
     When transport of ozone  and precursors  Into an  urban area 1s known
                                                               >i'6v
or suspected to be significant, air quality  measurements  aloft are desirable.
Instrumented aircraft equipped  with ozone  and  nitrogen oxide/n1trie  oxide
monitors are generally used.   Grab samples should also be collected in order
to characterize the organic species mix  aloft.   Aircraft measurements are
usually limited to days on which high ozone  levels  are expected.
C.   METEOROLOGICAL DATA
     As stated earlier, the Airshed Model  simulates the urban  atmosphere
under the constraint that pollutant mass 1s  conserved.  Conservation of
mass requires that pollutant transport through the faces of each grid cell
                                  53

-------
 be known  or  estimated.  This  Includes  both advectlon 1n the horizontal and
 dispersion 1n  the  vertical.   Advectlon 1s defined by the time-averaged wind
 vector while dispersion 1s related to  fluctuations 1n the wind and to vertical
 turbulence.  Variations 1n the vertical distribution of temperature affect
 the vertical extent of mixing within a column of grid cells.  The height
 to which  the mixed layer extends  1s termed the mixing height.  In order to
 simulate  all these processes, a three-dimensional wind field and a two-
 dimensional  field  of mixing heights, as well as atmospheric stability, all
 as a function  of time, are required.
     Another Important meteorological  parameter 1s solar Insolation since
 several key  photochemical reactions are driven by sunlight.  The amount of
 insolation present, as a function of time, must therefore be known or
 estimated.
     1.  Requirements for Meteorological Data
         Meteorological data  needed to characterize the wind field, mixing
 height, stability, and solar  Insolation are listed 1n Table III-5.  Hourly
 surface observations of wind  speed and wind direction are necessary to
 describe pollutant transport  at the surface while vertical profiles of
 speed and direction at various times during the day are necessary to
 describe the upper-level winds.   Measurements of vertical temperature
 gradients at various times are used together with hourly air temperatures
 at the surface to estimate hourly mixing heights.  Vertical temperature
 gradients are also used (along with solar Insolation data) to characterize
atmospheric stability within  the  mixed  layer.
     In order to adjust the photochemical rate constants so as to reflect
the amount of radiation available, solar Insolation measurements are
                                   54

-------
                               Table  III-5

                        Meteorological  Data  Needs
Wind Field                          Hourly surface observations  of
                                    wind speed  and direction  through-
                                    out region; vertical  profiles  of
                                    wind speed  and direction  and their
                                    temporal  variation.
Mixing Height and Stability         Hourly surface temperature
                                    observations throughout region;
                                    vertical  temperature gradients
                                    and their temporal variation.
Miscellaneous Data                  Hourly surface observations of
                                    solar Insolation; surface observa-
                                    tions of relative humidity and
                                    pressure and their temporal variation,
                                    55

-------
 needed.   Surface air temperatures  are  also  used  to adjust various chemical
 reaction  rate  constants.   Observations of relative humidity are needed 1n
 order  to  compute water vapor concentrations which 1n turn enter Into the
 photochemistry of oxldant  formation  (though the  effect  1s minor).  Con-
 version of pollutant concentrations  from a  mass  to a volume basis requires
 that atmospheric pressure  be known.
     2.   Collection  of Meteorological  Data
          The number  of meteorological  monitoring stations needed Is related
 to the size of the city and  the  complexity  of the wind  field.  Metropolitan
 areas  located  adjacent to  coastlines or lakeshores and  areas located 1n
 complex terrain will  require special consideration when designing the
 meteorological  monitoring  program.   As discussed 1n Chapter IV, Preparation
 of Model  Input Data,  wind  field  modeling may  be  required 1n such situations.
 The wind  data  collection effort  would  then  be directed  at satisfying the
 needs  of  the model chosen.   Even for Inland areas having relatively flat
 terrain,  significant heat  Island effects may  be  encountered.  Sufficient
 spatial resolution 1s required to characterize these effects.  Meteorological
 data collection  should occur over the  same  three month  period as the air
 quality monitoring.   The following discussion of requirements 1s applicable
 to metropolitan  areas with generally flat terrain and minimal sea breeze
 effects.
     In general, wind Instrumentation  should  be  collocated with the air
 quality monitoring stations  discussed  previously.  The  upwind, urban, and
 downwind nature of the regional  air  flow can  therefore  be characterized
and upwind-downwind trajectories can be determined.  However, care should
be taken to Insure that wind  Instruments are  sited to avoid mlcroscale
                                   56

-------
wind effects such as may be experienced  1n  the  lee of buildings or 1n close
proximity to topographic features.   Depending on city size and configu-
ration, as well as factors mentioned previously, the minimum  required
number of upwind stations may range from one to three,  urban  stations from
two to five, and downwind stations  from  two to  four 1n  number.  Wind
Instruments at existing air quality monitoring  stations should be
Incorporated 1n the study network 1f possible.  National Weather  Service
wind data, which 1s usually collected at major  airports and should be
available 1n most all metropolitan  areas, can be used to further  supplement
the wind monitoring network.
     Mixing heights are oftentimes  determined by measuring the  vertical
temperature gradient 1n the atmosphere.   Although  twice dally mixing heights
are routinely available from the National Weather  Service In  many metro-
politan areas, the measurements are normally taken 1n  rural  rather than urban
settings.  These data should therefore be supplemented  with  additional
radiosonde ascents and surface temperature measurements 1n order to charac-
terize mixing heights over the modeling region  and their variation with time.
In particular, estimates are needed of the minimum mixing height before
sunrise, the mixing height during the nrld-morning  when the rate of rise 1s
quite rapid, and the maximum mixing height 1n the  afternoon.   A minimum of
three ascents each day should be made at an urban  site with release times
of 0400, 1000, and 1500, for example.  Upwind of the city, the primary con-
cern Is the development of the mixing layer after sunrise.  Therefore,
radiosonde releases should also be conducted at an upwind site at 0500
and 0900, for example.  Alternatively, an acoustic sounder may be an
advantageous and economical method of obtaining morning mixing heights  at
                                    57

-------
 the upwind site.  (However, if an acoustic sounder  1s  used, an additional
 plbal site should be added to those suggested  below.)
      If aircraft flights are conducted to measure pollutant concentrations
 aloft, the aircraft should be equipped with temperature sensors to further
                                            t
 define the temperature structure of the atmosphere.  In addition, since
 surface temperature measurements CA" be used to  help Interpolate mixing
 heights temporally and spatially, such measurements should be taken at the
 wind monitoring stations discussed previously.
      The radiosonde ascents should be supplemented with plbal ascents 1n
 order to characterize air flow aloft.  By tracking the plbal or radiosonde
 over time, vertical  profiles of wind speed and wind direction can be
 Inferred.   Generally two or three plbal  release  sites  are desirable.
 The locations  and times  of their release should  be coordinated with the
 radiosonde ascents so as to provide a maximum overall  representation of
 the regional wind field.
      Solar Insolation (both direct and diffuse)  1s readily measured using a
 pyranometer.   The number of sites depends on the size  of the city, and
 the presence of water or terrain which might Influence cloud cover.  In
 general  however, two monitoring sites are desirable.   The Instruments should
 be  collocated  with the wind and air quality Instrumentation at stations
 representative of upwind,  urban, or downwind locations.
      In  regard to relative humidity and  atmospheric pressure, no additional
 data  need  normally be collected.   National  Weather Service or local agency
 surface  measurements,  normally  taken every hour, are generally sufficient for
modeling purposes.
     Figure III-4 shows  a meteorological  monitoring network for a major
metropolitan area  having a well-defined  predominant wind direction from
                                   58

-------
W8
WD
SR
T
RW
PB
Wind Speed
Wind Direction
Solar Radiation
Temperature
Rawlnnnde
Pibal
                                                                            ATLANTIC



                                                                             OCEAN
Figure 111-4. A Meteorological Monitoring Network for Philadelphia.
                                  59

-------
the west and southwest during periods of high ozone concentrations.   In
this example, the number of upwind, urban, and downwind surface sites 1s
2, 5, and 5 respectively.  Wind speed and wind direction are measured
at each station while temperature 1s recorded at selected sites upwind,
within, and downwind of the urban area.  Solar radiation 1s measured
both upwind and downwind.  Winds aloft are measured at three sites,  one
each at an upwind, urban, and downwind location*  Temperature soundings
are also taken at the urban site.  The number of meteorological monitoring
sites for any particular city will depend on the nature and complexity
of the wind field and of the surrounding terrain.
                                   60

-------
             CHAPTER IV.  PREPARATION OF  MODEL INPUT  DATA
     Once the emissions, air quality, and meteorological data  described
in Chapter III have been assembled, the data  must be  prepared  for model
input.  Data preparation includes selection of a  set  of individual modeling
days from the meteorological and air quality  data bases for which model
input data is to be prepared.  Once the data  has  been prepared,  the  model
can be run.  Model operation is described in  detail in the User's Manual
           u
for the Airshed Model  (Ames, et. al., 1978)  and  will not  be discussed here.
Computer resources necessary to run the model are, however, described
briefly in Chapter VI, Resource Requirements  for  an Airshed Model Appli-
cation.
A.   DAY SELECTION CRITERIA
     Prior to attempting to prepare the aerometric data for model input, one
must select the days to be modeled.  Normally, three  or four days should be
selected.  Several selection criteria should  be observed.   First, ozone
levels should be near the maximum that occurred during the air quality
monitoring program.  Such high ozone episodes often occur  over a two or
three-day period.  Second, weekend days should be avoided.  Use of an
emission Inventory developed for a weekday may Introduce errors if used on a
weekend day.  Total regional emissions are lower  on weekends and the temporal
and spatial distribution of the emissions are different.  Third, the
      At this time, only a draft has been provided by Systems Applications,
Inc.  The manual will be available for public release upon completion of a
finalized version.
                                   61

-------
 aerometHc  data  base  should  be relatively complete.  Excessive occurrences
 of missing  data  makes data preparation difficult and may lead to a poor or
 suspect  verification.  Fourth, the air quality and meteorological data should
 not exhibit abnormal  complexity.  Unusual wind fields or concentration
 gradients may  place undue stress on the model.  Such occurrences may cause
 the data itself  to be suspect in some cases.  Passage of warm or cold fronts
 may introduce  unacceptable complexities.  Under no circumstances should any
 precipitation  have occurred  during the pre-dawn to mid-afternoon hours.  In
 general, sunny skies  and light winds associated with slow-moving high pressure
 systems  are conducive to high ozone levels.

 B.  DATA INPUT FILES
     the Airshed Model  contains eleven data preparation programs which assist
 the user in creating  the aerometric and emissions input files required for
 a simulation run.  Figure IV-1 shows the input data files while Table IV-1
 provides a  short description of each.  A variety of alternate methods for
 preparing the  aerometric data are possible.  Selection of a method depends
 on the nature  of the  data base.  The level of sophistication of the method
 selected follows from the level of sophistication of the data base.  Several
methods  depend upon Interpolation/extrapolation routines 1n the horizontal
 using station  measurements.  Vertical resolution 1s generally specified by
means of normalized vertical profiles.  Different methods may be used for
 preparing the  data for different subregions.  This allows the modeler to
better utilize a spatially uneven data base as well as characterize spatial
differences more realistically.  Table IV-2 Indicates the spatial and
temporal  resolution for each data input parameter that is achieved by
utilizing the preparation programs.
                                   62

-------
                  METEOROLOGICAL DATA
                                                  AIR QUALITY DATA
       ri
WIND
              DIFFBREAK
ri
METSCALARS
CO
              REGION TOP
AIR QUALITY
                   TEMPERATUR
                    TERRAIN
                                         BOUNDARY  '
                                                                      TOPCONC
                                                       I
                                                      AIRSHED
                                                     SIMULATION
                                                     PROGRAM
                                Figure IV-1.  Input data files for airshed simulation program.
                                                                                                  EMISSIONS DATA
                                                          ri
                                 EMISSIONS
                                                                         PTSOURCE

-------
                             Table  IV-1.
       DESCRIPTION OF THE  INPUT FILES TO THE AIRSHED MODEL
DIFFBREAK       This file contains the mixing height for each column
                of cells at the beginning and end of each hour of the
                simulation.
     *
REGIONTOP       This file contains the height of each column of cells at
                the beginning and end of each hour of the simulation.  If
                this height Is greater than the mixing height, the cell or
                cells above the mixing height are assumed to be within an
                Inversion.

WIND'           This file contains the x and y components of the wind
                velocity for every grid cell for each hour of the simu-
                lation.  Also the maximum wind speed for the entire grid
                and average wind speeds at each boundary for each hour
                are Included on this file.

METSCALARS      This file contains the hourly values of the meteorological
                parameters that do not vary spatially.  These scalars are
                the N02 photolysis rate constant, the concentration of
                water vapor, the temperature gradient above and below the
                Inversion base, the atmospheric pressure, and the exposure
                class.

                This file contains the Initial concentrations of each
                species for each grid cell at the start of the simulation.
                          \
                This file contains the location of the modeling region
                boundaries!  This file also contains the concentration of
                each species that 1s used as the boundary condition along
                each boundary segment at each vertical level.

TOPCONC         This file contains the concentration of each species for
                the area above the modeling region.  These concentrations
                are the boundary conditions for vertical Integration.

TEMPERATUR      This file contains the hourly temperature for each surface
                layer grid cell.
                           •
EMISSIONS       This file contains the ground level emissions of NO, N02,
                five carbon bond categories, and CO for each grid square
                for each hour of the simulation.
AIRQUALITY


BOUNDARY
                                     64

-------
Table IV-1.  (Continued)


PTSOURCE       This file contains the point source Information,
               Including the stack height, temperature and flow
               rate, the plume rise, the grid cell Into which the
               emissions are emitted, and the emissions rates for
               NO, N02, five carbon bond categories, and CO for
               each point source for each hour.

TERRAIN        This file contains the value of the surface roughness
               and deposition factor for each grid square.
                                        65

-------
                                  Table  IV-2

             DATA-RELATED  INPUT  PARAMETERS FOR THE AIRSHED MODEL
 Description

 METEOROLOGY

 Horizontal (u-v)
 winds  (m/sec)
 Reference  height of
 surface wind monitoring
 stations (m)

 Diffusion  break (m)
Spatial and
Temporal
Resolution*
     XJJZ   t
Top of modeling           x
region (m)

Ground-level tempera-     x
tures (°C)

Surface atmospheric
pressure  (mb)

Temperature gradient below
diffusion break (°C/m)

Temperature gradient above
diffusion break (°C/m)

Water concentration 1n
the atmosphere (ppm)
                          >
Exposure (stability class)

Radiation Intensity
factor (per m1n)

Surface roughness (cm)    x
Remarks
               The vertical component, w, 1s computed
               by the Airshed Model, rendering the
               resultant wind field  mass consistent

               Used 1n the d1ffus1v1ty algorithm
           x   Elevation at which the stability
               structure of the atmosphere changes
               markedly (e.g., an Inversion or
               thermal Internal boundary layer)

           x   Generally defined 1n relation to
               diffusion break

           x   Used 1n kinetic module
           x  .Used 1n computing grid cell  concen-
               trations associated with emissions

           x   Used in plume rise calculations


           x   Used In plume rise calculations


           x   Used 1n kinetic module


           x   Uspd 1n dlffuslvlty algorithm

           x   Used 1n kinetic module
               Used in dlffuslvlty and surface
               sink algorithms
                                   66

-------
Table IV-2 (continued)
Description

Vegetation factor

AIR QUALITY

Initial conditions
(pphm)

Boundary conditions
(pphm)

Concentrations above
top of modeling
region (pphm)

EMISSIONS

Lumped ground-level
emissions from
stationary and mobile
sources (gm-moles/hr)

Elevated stationary
point source emissions
(gm-moles/hr)

Elevated point source
stack data Including
height (m) temperature
(°K) diameter (m) and
velocity (m/sec)
                         Spatial and
                         Temporal
                         Resolution*
x    x
x    x
x    x
                       Remarks

          Used 1n surface sink algorithms
Required for NO, N02, 0,, five carbon
bond categories, PAN, BZA, and CO

Required for same pollutants as
above

Required for same pollutants as
above
          Required for NO, N02, five carbon-
          bond categories and CO
Emissions from tall stacks  for
the above pollutants are
required

Used in plume rise calculations
 xy  Indicates two dimensional data field
 xyz Indicates three ^Pfinensiohal data field
   t indicates time varying data field
                                   67

-------
      1.   Preparing  the  Meteorological  Data
          Preparation  of the METSCALAR  and TEMPERATUR input files is relatively
 straightforward.  The METSCALARS file  contains those meteorological parameters
 which do  not  vary spatially.  Hourly values of the average temperature gradi-
 ent below and above the mixing height  and hourly values of atmospheric pres-
 sure, water concentration  (derived from humidity measurements), and exposure
 (stability) class must  be  specified.   In addition, hourly N02 photolysis rate
 constants must be input.   The latter are readily calculated using measured
 solar insolation and  estimated solar angle (Schere and Demerjian, 1978).
 Preparation of the  TEMPERATUR file requires that hourly station temperature
 readings  must be input  and the desired interpolation routine selected.
          Preparation  of the DIFFBREAK  file generally requires the use of
 algorithms or techniques external to the data preparation program.  Vertical
 temperature gradient  readings (2 to 3  sites, 3 to 4 times daily) and surface
 temperature readings  (5 to 12 sites, hourly) are used to determine hourly
mixing heights at all stations (7 to 15 sites).  These values are then input
 to the data preparation program and a  suitable spatial interpolation routine
 is selected.  Creation  of  the REGIONTOP file does not involve any actual
 data  input.   Rather,  the user generally specifies the number and height of
grid  cells above the  mixing height.  Alternatively, a fixed region height
may be used.
          Preparation  of the WIND file  requires special expertise in wind
field modeling and analysis and the use of specialized wind generation
programs.  Such programs are not provided by the Airshed Model.  The WIND
file should be recognized  as a critical element of satisfactory model per-
formance.  For these  reasons, the WIND file is undoubtedly the most
                                   68

-------
challenging file to prepare.   An  Iterative  procedure, whereby wind field
modeling 1s conducted and the resulting wind  field 1s evaluated, 1s usually
necessary.
     Three distinct approaches toward  generating wind field Inputs may be
taken.  The simplest approach Is  the use of an Interpolation scheme whereby
station readings of wind speed and wind direction are used to approximate
values 1n each grid cell.  Vertical wind profiles derived from plbal,
rawlnsonde, or other measurements, are also used.  The  Interpolation scheme
must be followed by a smoothing of the resulting multilayer, horizontal
wind fields.  This 1s necessary to reduce divergence in the resulting wind
field.  The Airshed Model simulation program  Itself then removes any remain-
Ing divergence by generating compensating vertical winds, thus Insuring  a  mass
consistent wind field.
     An alternative and more sophisticated  approach 1s  the  use of  so-called
"diagnostic models."  Such models are  appropriate where the local  terrain  con-
figuration has a significant effect  on wind patterns.   These models, which
Incorporate some of the same concepts  discussed  above,  use  a basic conserva-
tion of mass equation subject to  various  boundary constraints.   Such  con-
straints may Include heat Island  effects  as well as terrain blocking  and
channeling effects and Inversion  limiting effects.
     The most complex approach 1s a  dynamic wind field  model  based on the
solution of mass, momentum, and energy equations.   These models may be needed
In areas with significant thermally generated sea  or  lake breezes.  Unfortu-
nately, their data Input requirements  are considerable and their computer
requirements are comparable with those of the Airshed Model Itself.
     A technique, which 1s available as an Airshed Model data preparation
program, may be used in special situations where terrain and water effects
                                   69

-------
are negligible and little wind data 1s available.  It uses an average wind
vector together with a network of station temperature readings to approximate
urban heat island effects.
     Preparation of the TERRAIN file involves the estimation of surface
roughness and surface deposition factors throughout the modeling region.
Pollutant uptake by water bodies and vegetative surfaces may influence
ozone concentrations.  Standard factors are used to represent uptake by
various types of surfaces such as trees, grasses, water, and man-made mate-
rials.  The relative composition of individual grid squares and/or subregional
areas in terms of these surface types can be approximated using land-use  maps
and assumed mixes for each land use category.  Surface roughness 1s employed
by the model to characterize vertical dispersion near the ground.  Land-use
maps, topographic maps, and the surface types described above are used to
determine the roughness controlling features so standard roughness length
values can be selected.  As with surface deposition, roughness lengths may
be specified at the subregional level or at the grid square level depending
on Its spatial variability.

   2.  Preparing the A1r Quality Data
         Three files use air quality data:  BOUNDARY. AIRQUALITY, and TOPCONC.
Although each 1s created by a separate data preparation program, care should
be exercised to assure the files are mutually consistent.  For example, the
vertical distribution of upwind boundary conditions should be consistent  with
the values used above the modeling region.  Otherwise, an artificial disconti-
nuity is introduced.  Each file contains values for the eleven pollutants
listed in Table IV-3.
                                   70

-------
                                  Table IV-3
                   POLLUTANTS TREATED BY THE AIRSHED MODEL

Pollutant                                          Airshed Model Name
Ozone                                                    03
Nitric Oxide                                             NO
Nitrogen Dioxide                                         N02
Organlcs:
    Highly Reactive Double Bonds                         OLE
    Aromatic Rings                                       ARO
    Single-Bonded Carbon Atoms                           PAR
    Carbonyl Bonds                                       CARB
    Moderately Reactive Dbuble Bonds                     ETH
Carbon Monoxide                                          CO
Benzaldehyde                                             BZA
Peroxyacetyl Nitrate                                     PAN
                                         71

-------
     The AIRQUALITY file contains the initial concentration of each pollutant
 1n each grid cell at the start of the simulation.  Prior to use of the prepara-
 ration program, nonmethane hydrocarbon values at the upwind and urban sta-
 tions for the first hour are converted Into the five carbon bond categories
 using the organic species data from early morning grab samples.  (This con-
 version step 1s discussed further in regard to emissions in the Appendix,
 Technical Description of the Airshed Model.)  Then these values, along with
 values of ozone, nitric oxide, and nitrogen dioxide are specified for each
 station at which they were measured.  A suitable (horizontal) Interpolation
 routine 1s then selected, one for each pollutant.  If aircraft spirals were
 used to collect upper-level early morning air quality data, a vertical pro-
 file for each pollutant and the location the spiral was flown 1s specified.
 If not, a simple vertical profile must nevertheless by chosen.  For example,
 values at the ground might be extended up to the mixing height.  Estimated
 values aloft (discussed below) could be used above the mixing height.  When
 Interpolating horizontally or specifying a vertical profile, the modeling
 region can be broken up Into subreglons.  However, designation of subregions
 must be observed for all pollutants.
     For carbon monoxide interpolation routines may be used with one-hour
 data from the existing monitoring network.  Without data aloft, values
 could be assumed constant 1n the vertical.  For those pollutants not mea-
 sured during the monitoring program (BZA and PAN), typical reference values
may be used (Ames, et. al., 1978).
     The BOUNDARY file contains the hourly concentrations of each pollutant
for each boundary cell.  The region boundaries must be defined first, how-
ever.  This 1s done by assigning Hne segments which completely enclose the
                                    72

-------
region.  A ground-level  value for each  pollutant 1s then Input for each
boundary segment each hour.   Deriving these values from the air quality data
base 1s mostly a matter  of judgment.  The prevailing wind direction each hour
1s used to determine which stations are upwind or downwind.  Downwind boundary
values are not Important although they  must be input.  As for Initial con-
ditions, nonmethane hydrocarbons must be converted to five carbon bond
categories.  A normalized vertical profile must also be specified for each
boundary segment for each pollutant.  These can be varied through the day.
The vertical profiles should be tied to the ground-level value and the value
aloft.  If aircraft data has been collected, profile preparation Involves
selecting suitable profile methods and  preparing the profile data sets so  as
to make the best use of  available data. Even 1f upper-level data have not
been collected, a simple assumed profile must still be entered.
     The TOPCONC file contains the hourly concentrations of each pollutant
above the modeling region.  If aircraft data have been collected the data
are used to determine the concentrations aloft of ozone, nitrogen dioxide,
and nitric oxide.  The results of aircraft organic species grab  samples  are
used to estimate the concentrations of  the five carbon-bond categories.   If the
data show any significant spatial variation, this should be represented  either
by using an Interpolation routine with  so-called  "station"  readings or by
specifying a value for each  grid square across the top of the region.  Sig-
nificant temporal variations should  also be  taken Into  account.   If no upper-
level air quality data are available,  values aloft may  be estimated crudely
by using the late morning air quality measurements  at an upwind ground-level
monitoring station.  This method  assumes that  upwind ground-level  concentra-
tions are Indicative of concentrations aloft upon dissipation of the nocturnal
                                  73

-------
surface Inversion.  For those pollutants not measured either upwind or aloft
during the monitoring program (CO, BZA, or PAN) typical reference values may
be used (Ames, et. a., 1978).
     3.  Preparing the Emissions Data
         As discussed 1n Chapter III, Data Needs, the emissions data base
should consist of hourly grldded ground-level emissions from minor point sources,
area sources, and mobile sources.  Each of these three files should contain the
emissions for eight pollutants:  NO, N02, OLE, PAR, CARB, ARO, ETH, and CO.
These three files must then be merged to create the EMISSIONS file required
by the Airshed Simulation Program.
          While the EMISSIONS file contains the ground-level emissions, the
elevated point source emissions are contained on the PTSOURCE file.  Prepara-
tion of this file involves specifying the location and stack parameters
(height, diameter, velocity, and temperature) for each major point source,
entering the hourly emissions of the eight pollutants for each source, and
running the data preparation program provided.  The data preparation program
utilizes the meteorological Input files described earlier to compute plume
rise and then assign the point source emissions to individual elevated
grid cells.
                                   74

-------
              CHAPTER V.   EVALUATION OF  MODEL  PERFORMANCE
     Before a model  1s used for assessing alternate  control measures  for
reducing ozone precursor  emissions,  one  must demonstrate the  ability  of
the model to estimate ozone concentrations satisfactorily.  Such  a  veri-
fication on several  high  ozone days  1s desirable.  By  simulating  several
meteorological and air quality situations* model performance  can  be
evaluated over a range of conditions conducive to  ozone formation.  Should
verification attempts fall on one day, other days  are  stm available for
analysis.  Also, control  measures may be more  or less  effective depending
on the level of pollutant transport  and  the mesoscale  meteorological
conditions.  In addition, verification on multiple days allows  control
strategies to be tested under different  conditions.
     Normally, single day verifications  should be  conducted for at  least  14
hours, usually from about 5:00 a.m.  to about 7:00  p.m. Such  a  time span
covers the morning emissions peak and the afternoon  ozone  peak  should it
occur late 1n the day.  However, 1f  downwind ozone monitors clearly show  that
ozone levels drop off much earlier 1n the day, the length  of  the  simulation
may be shortened.
     A number of performance measures may be used  to evaluate whether or  not
the model 1s performing satisfactorily.   Normally, performance  1s evaluated
separately on each verification day. Both graphical and  statistical  tech-
niques are useful.  Several techniques  should  be chosen because different ones
measure different aspects of the model's performance.   Some measure the model's
performance at estimating peak concentrations, others measure spatial and temporal
correlations, while still others measure the overall accuracy of the model results.
     The U. S. Environmental Protection Agency has recently published two
reports addressing the evaluation of model performance:  Procedures for
                                  75

-------
Evaluating the Performance of Air Quality Simulation Models (Hlllyer,
Reynolds, and Roth, 1979) and Performance Measures and Standards for Air
Quality Simulation Models (Hayes, 1979).  The first report summarizes
various issues and considerations pertaining to model evaluation studies.
The second report, in part, presents a number of specific performance  measures.
     Comparisons of model-estimated concentrations and observed concentrations
must be done cautiously.  Obviously, Instrument error 1s always a complicating
factor.  Beyond this, however, the observed value at best represents a con-
centration measured at a single point.  In contrast, the estimated value
represents an average concentration within a surface layer grid cell;  gradients
within a cell cannot be accounted for.  The Airshed Model does perform an
east-west and a north-south linear interpolation between the four closest
grid cells to obtain a point estimate.  The grid cell average value 1s used
to represent the value at the centroid of the cell for the interpolation.
This procedure does not, however, compensate for the fundamental difference
between observed and estimated values.  It only provides an objective
method for comparing model results to measured concentrations.  Clearly, the
representativeness of the monitor site is a key factor.  For example, model
estimates of NO  or CO that are much lower than monitor readings while at
the same time ozone estimates that are significantly higher may indicate the
monitor is being unduly affected by localized emission sources under certain
meteorological conditions.  These kinds of disparities point up the need for
properly siting ozone monitoring equipment when setting up the ambient
network discussed in Chapter III.  In general, the observed concentrations
are nevertheless assumed to be the "true values" when using performance
measures to evaluate model results.
                                  76

-------
     The following discussion presents  a  variety of performance measures,
both graphical  and statistical,  which have  been  Identified  for use  in
evaluating model  performance. The measures discussed  here  are not  Intended
as recommendations, as further experience with a wide  variety of measures
in various applications 1s needed in this newly  developing  field.
A.   GRAPHICAL TECHNIQUES
     Scatter diagrams are one of the most common graphical  methods  of  com-
paring observed and estimated concentrations.  Normally  the observed values  are
plotted on the abscissa on the same scale as the model estimates which are
plotted on the ordlnate.  A 45 degree 11ne  through  the origin then  represents
a perfect correlation.  The distance of points from this line 1s  a  measure of
the deviation while the clustering of points above  or  below is a  measure of
bias.  As shown in Figure V-1 maximum values at  each station, mean  values
over a certain time Interval (or the entire simulation), or all  hourly values
at all stations may be visually  compared  in this way.
     Time histories of observed  versus  estimated concentrations  at  individ-
ual stations may also be presented graphically as shown  in  Figure V-2.
Such plots show whether the model estimates conform to the  temporal pat-
terns exhibited by the monitoring data  and  the extent  to which  model  esti-
mates are out of phase with actual values.
     If the monitoring network is sufficiently dense,  it may be possible to
plot isopleths of the observed concentrations  and compare these to isopleths
of the model estimates, plotted  at the same scale, as  shown in Figure V-3.
Two attributes may be examined using such plots:  general pattern corres-
pondence and directional alignment.  General pattern correspondence is more
important than simple directional alignment.  Nonalignment may be caused
                                    77

-------
                     9- 5
                            a.
                           0       5      10     15
                           Observed concentrations (pphm)
                14
                1.
                1
                .12
                    _ b.
                        i     i     i     I     r
                                     DSL.
   DRCO
   DBUo
                        I     !    I     I     I
                         oOLI   -
DRMo.
                     DSFo
                             2345     67

                              Observed mean (pphm)
                i8
                1
                a.
                i

                1
                                                  I
                          4       8      12       16
                             Observed maximum (pphm)
Figure  V-l  Scatter diagrams of (a) observed oxidant concentrations versus
calculated hourly average ozone concentrations, (b) observed versus  calculated
station mean concentrations, and (c) observed versus calculated station maximum
concentrations (Duewer, MacCracken, and Walton, 1978).
                                      78

-------
                                                      -OBSERVED
                                                  —-«-fltEO!CTED
                                           »-0.—A—.0   PARKER RO
 I  f  i   f  ft  ff  tt  ¥  1   I   i  i   t   »
               T
-------
00
o
                             ESTIMATED
                                                                                          OBSERVED
                          Figure V- 31. Comparison of estimated and observed ozone isopleths for the hour

                          of maximum observed concentration.

-------
by small errors 1n the wind field while a  lack of pattern  correspondence
Indicates a more fundamental  problem with  the data Inputs  or  the model
Itself.  Plots such as Figure V-3 may be further Interpreted  by comparing
the areas over which ozone levels exceed certain threshold values  above the
standard.
B.   STATISTICAL TECHNIQUES
     As shown 1n Table V-l, a variety of numerical and statistical
measures may be used for evaluating model  performance.  It 1s Important to
note at the outset that 1n most applications, the number of days modeled 1s
Insufficient to make any statistical statement regarding the  ability of the
model to estimate peak regional oxldant concentrations. Therefore 1t 1s not
possible to say how the model might be expected to perform on a day other
than those for which a verification analysis 1s actually undertaken.
         The ability of the model to accurately estimate peak ozone concen-
trations 1s crucial to the application of  the model for SIP planning pur-
poses.  The most obvious numerical measure 1s the difference between the
estimated maximum ozone concentration at any station and the observed
maximum ozone concentration at any station.  This comparison 1s made Irre-
spective of the time of day during which the maxima occurred.  Another
numerical measure of peak performance 1s derived by taking the ratio of
the estimated maximum to the observed maximum concentration, one  ratio for
each station and then finding the median ratio.  The median rather than
the average value is used to avoid unnecessary bias that might be caused
by a bad estimate at one or two stations,  particularly at locations where
both values are low.
                                    81

-------
     Table V-1.  Some Numerical and Statistical Performance Measures

Performance Attribute                          Performance Measures
Accuracy of Peak
Estimates
Overall Accuracy
of Estimates
Absence of Bias
Diurnal Pattern
Spatial Pattern
Difference between estimated and observed
maximum concentration regardless of time
or location

Median ratio of estimated to observed
maximum station concentrations

Median relative absolute deviation between
estimated and observed maximum station
concentrations

Correlation coefficient of estimated and
observed maximum station concentrations

Median relative absolute deviation between
estimated and observed station concen-
trations taken over all stations and all
hours

Average absolute deviation and standard
deviation of absolute deviations between
estimated and observed station concen-
trations taken over all stations and all
hours

Difference 1n the frequency of occurrence
of values above some threshold, between
estimated and observed station concentra-
tions considering all stations and all hours

Correlation coefficient of estimated and
observed station concentrations taken over
all stations and all hours

Relative deviation between estimated and
observed maximum concentrations averaged
over all stations

Relative deviation between estimated and
observed station concentrations averaged
over all stations for hours when the
observed value exceeds some threshold

Temporal correlation coefficient qf
estimated and observed station concentra-
tions averaged over all stations

Spatial correlation coefficient of
estimated and observed station concentra-
tions averaged over all hours
                                 82

-------
     The median relative absolute deviation of maximum values,  taken  over
all stations, times 100, 1s a good measure of the percent  error associated
with the peak ozone model estimate.  The relative absolute deviation
1s defined here as the absolute value of the difference between the esti-
mated and observed values, divided by the observed value.   Another good
statistical measure of peak model performance 1s  the  correlation coefficient,
taken over all estimated/observed maximum concentration pairs,  one pair  for
each station.
     Because the Intent of control  strategy analysis  1s to reduce ozone
concentrations below the standard and to reduce population exposure,  it  1s
also Important that the model perform satisfactorily  over  a wide range of
ozone concentrations.  As with maximum concentrations, the median relative
absolute deviation 1s also a measure of the overall percent error of  the
model estimates.  In this case, however, rather than  limiting the calculation
to the maximum values, the median 1s taken using  the  entire set of deviations
for all hours during the simulation.  The overall correlation of all  esti-
mated and observed concentration pairs may also be computed. Measures of
the absolute error and Its variability are the average absolute deviation
and the standard deviation of absolute deviations, respectively.  The latter
statistic Involves the computation of the deviation of the absolute  deviation
of an estimated/observed pair from the average absolute deviation of all
such pairs.  Both statistics are computed over all stations and over all
hours of the simulation.
     A somewhat different performance measure Involves the comparison of
the cumulative frequency distribution of estimated ozone concentrations
to the cumulative frequency distribution of observed  ozone concentrations,
                                   83

-------
again taken over all stations and all hours as 1n Figure V-4.  A numeric
comparison 1s then made of the frequency with which various concentration
levels are exceeded.
     One should also evaluate the model results for systematic bias.  Bias
occurs when the model has a tendency toward repeated errors 1n the same
direction, either over-estimation or underestimation.  In this case, the
algebraic average deviation rather than the absolute average 1s used.
Normalizing the deviation by the observed concentration before averaging
gives the average relative deviation, a measure on a fractional basis of
the seriousness of the bias over any range of concentrations.  Possible
variations of this measure Include restricting the computation to the maximum
values for each station or to the hours during which the observed concen-
tration exceeds some value, such as the ambient air quality standard.  The
overall bias may be assessed by extending the computation to Include all
stations and all simulation hours.
     Quantitative statistical measures may also be used to evaluate the
model's ability to generate satisfactory diurnal and spatial patterns.  The
temporal correlation coefficient, averaged over all stations, and the spatial
correlation coefficient, averaged over all hours, are two such Indices.
Variations Involve stratifying the data so that, for example, the temporal
correlation 1s determined for stations Individually or for groups of sta-
tions.  Similarly, spatial correlations may be taken over one or more hours,
usually the peak period.
     C.  REVERIFICATION
         In the event of poor model  performance, one must determine what
1s causing the model  estimates to be unsatisfactory.  The possible causes
                                  84

-------
00
01
             40



             30





             20
S    10

5    9
£    8

g    7

2    6
             111    I   I    I     I     I—I   I   I   I   I    I    I      II      I   I   I    I
                      - -O- — OBSERVED

                               ESTIMATED
                                  279 DATA PAIRS FROM 3 DAYS. 14 HOURS, 9 STATIONS
                     I  I  I   I   I    I     I     V    1  I   I   I   I    I    I	LJ	I   I   I   I  I
                      0.1 0.2      12     5   10     20   30  40  50  60  70  80    90   95    98  99  99.5 99.8 99.9

                        ''                              CUMULATIVE FREQUENCY


                 Figure  V-4. Cumulative frequency distribution of estimated and observed ozone concentrations (adapted

                 from Hayes, 1979).

-------
 may be  grouped  Into  three areas:   (1) the emissions, air quality, or mete-
 orological  data bases may not  adequately or properly characterize conditions
 on  the  day  modeled;  (2)  the methods  used for preparing the air quality and
 meteorological  data  for  model  Input  may not be appropriate or were exercised
 Improperly; (3) the  model may  not  be suitable or does not simulate the
 atmospheric phenomena Important for  the day or location being modeled.  Unless
 a research  and  development type effort 1s being contemplated, one should
 probably not subject the last  cause  to any prolonged Investigation.  Rather,
 efforts should  be placed on Investigating the data base and Improving and/or
 correcting  the  methods by which the  data base was prepared for model Input.
     Generally, Investigation  of model performance problems should center
 Initially on the procedures used for preparing the aerometrlc data for model
 Input.  This Is because  the modeler  has the most control over this portion of
 the modeling study.  Invariably, deficiencies exist 1n the data base which
 necessitate assumptions  1n order to  prepare the data for the model.  Such
 assumptions  should be reevaluated.   For example, preparation of the wind field
 and mixing  height data are prime candidates for review.  These two Inputs are
                                                                         *,,
 primary driving forces in the model  simulation while the number of data points
 available and the methods used for their preparation methods may be less than
 Ideal.  Further analysis of the Input data may Improve model performance.
     When Investigating  the data base, attention 1s most often focused
 on the emission Inventory.  This 1s  due to the fact that emission Inventories
 are frequently found to be deficient 1n one or more respects.  One may find
 for example, that significant sources of hydrocarbons have been overlooked.
One may also find that not all  the vehicle traffic has been accounted for
or that the Input data used to compute mobile source emission factors 1s

                                    86

-------
not representative or sufficiently resolved spatially and temporally.
Another candidate for Investigation Is  the organic species data used for
allocating hydrocarbon emissions  to the five carbon-bond categories.  One
might find for example that serious errors 1n judgment have been made 1n
applying an Imperfect organic species data base to specific sources 1n
the Inventory.
     The modeler, due to his familiarity with the data base, the various
methodologies, and their strengths and  weaknesses should have a good Idea
of what areas warrant Investigation and modification.  Nevertheless, a
                                                                      >
knowledge of the sensitivity of the model to changes in Input data 1s
Invaluable.  For example, little  1s gained from expending much effort on
Devaluating stability class data If the model results are not sensitive
to stability class.  While no comprehensive sensitivity study of the Airshed
Model has been reported, Table V-2 summarizes the results of a number of
sensitivity analyses performed with previous versions of the model and
another photochemical grid model, the Llvermore Regional A1r Quality  (LIRAQ)
model.  If all reasonable and objective modifications to the Input data fall
to upgrade model performance to a satisfactory level, then  one  1s  forced  to
admit that the model 1s unable to simulate the atmospheric  conditions  existing
on that particular modeling day.   In such cases,  1t  1s  especially Important
that modeling of several different high ozone days has  been provided for
previously.
     In summary then, model verification 1s  not  in any way a "push-button"
operation.  Rather, it 1s an Iterative and  an  Incremental  process.  With
each execution of the simulation  program model  performance 1s evaluated,
problems are diagnosed, the Input data 1s reanalyzed, and the Input files are
modified until a satisfactory verification 1s achieved.
                                  87

-------
                                thrill,  £; *TnnS+Try  f ?e;?1tivi;^  of Selected  Photochemical  Grid
                                Models  to  Input  Data  Variations  (Reynolds, Tesche,  and  Reid,  1978)
                    Study Group
     Nodel Version
     anA Attributes
              HacCracken and
              Sauter (I97S)
00
00
              Deaerjlan (1976)
              Liu tt •!.  (1976a)
 LIRAQ photochemical  model

   Two-dimensional  time-
   dependent grid Mxfel

   LiMped kinetic mech-
   anism siMilar to
   Hecht-Selnfeld-Dodge
   Mechanism

   Mass-conserving  Mind
   field
SAI photochemical codel:
1973 version
SAI photochemical Mdel:
1973 version
  25 x 25 x 6 grid

  IS step Hecht-Selnfeld-
  Dodge kinetics

  Price numerical Method

  Empirical diffusion
  algorithm

  Two-dimensional Mind
  field

  Two-dimensional
  Initial conditions
  Sensitivity Analysis
       Variations

Relative humidity Mas
reduced fro» 40* by 20X

Nominal temperature Mas
Increased fro* 285°* to
304*K

Light Intensity Mas re-
duced by SOI
                                                                      Light Intensity Mas
                                                                      Increased by a factor
                                                                      of 2
                                                                      Initial hydrocarbons Mere
                                                                      Increased by a factor of 2
Initial NO; concentrations
were Increased by a  factor
of 2

Boundary conditions  were
reduced by SOX
Initial and boundary con-
ditions were reduced by
SOX

Wind directions were
randmly oerturbed by
0 or *22.5«
                                                                     Hind speeds wen randomly
                                                                     perturbed by 0 or ±1  mpti
                                                                      Hind station Measure-
                                                                      ments were:
                                                                        Increased SOS

                                                                        Increased 2SX

                                                                        Decreased 25X

                                                                        D*er««ad SOS
 Influence on Model  Predictions

 Peak ozone Increased by 31
 and peak NO; decreased by «J

 Peak ozone decreased by 21.
 and peak NO; Increased by 5X
Peak ozone decreased by 70S.
and NO; peak Magnitude rewined
unchanged but was delayed 4
hours

Peak ozone Increased by 100S.
and NO; peak Magnitude slight-
ly Increased and preceded base
case peak by 1-3/4 hours

NO; peak Increased by 61 and
was delayed approximately I
hour; ozone peak was not re-
ported, but the Increase in
ozone concentrations was delayed
by up to 3 hours

NO; peak Increased by 101 and
was delayed slightly! 03 re-
Mined unchanged

•Minor" differences occurred
In ozone prediction In the
eastern and northern portions
of the L.A.  basin; "significant"
differences were observed In the
western and central portions of
the basin

Predicted ozone In the northern
and eastern edges of basin were
reduced 20 to 30X

A 6.9S average deviation for
•mually prepared and 4.91
for automatically prepared
wind fields (based on CO
predictions) resulted

A 4.91 average deviation for Man-
ually prepared and 2.6X for auto-
matically prepared Mind fields
(based on CO predictions)  resulted

HaxlMUM absolute deviation from
the base case results for CO
were-.

  19.61

  11.BX

  20.2%

  Sl.TJ
                                                                                                                                               Remarks
LIRAQ sensitivity runs focused
on the kinetic module; accord-
ingly, sensitivity results are
more reflective of SMDQ charter
simulations  than urban airshed
simulations.
 In the automatic Mind field
 studies, perturbations were
 •ade to the Monitoring station
 measurements and then automatic
 procedures were employed to
 derive gridded wind fields.   In
 the Manual wind field cases.
 perturbations were Made to the
 gridded wind fields after they
 had been prepared Manually.

 The response of the Model to
 variations In wind speed varies
 with each chemical species and
 is time dependent.

-------
                                                              Table  *V-2 (Continued)
               Study Croup
Model Version
and. Attributes
   Sensitivity Analysis
	Variations

Horizontal diffusion MS
decreased to 0 and  In-
creased to 500 mVs
                                                                  Vertical  dlffuslvlty MS
                                                                  decreased to 0.5 m'/ssc
                                                                  and Increased to 50 mz/sec
                                                                  Nixing depths were In-
                                                                  creased and decreased by
                                                                  251
oo
vo
                                                                  Radiation Intensity MS
                                                                  Increased and decreased
                                                                  by 30%
                                                                  (Missions rate (ground
                                                                  based) MS Increased and
                                                                  decreased by 151
Influence on Model Predictions

For Ku • 0. the Maxima abso-
lute deviation for CO ranged
between 0.52 and 2.021 fron.
0600 to 1600 hours

For KH • 500 m2/sec. the
maximum absolute deviation for
CO ranged between 4.4 and 12.9%
from 0600 to 1600 hours

The effect of varying vertical
dlffuslvlty by an order of Mag-
nitude MS about the same as
that of varying the wind speed
by 25 to 50*

MaximiM absolute percentage
deviations for the Increase and
decrease, respectively. Mere:

  For CO. Stand 121

  For HO. 11% and 18.51

  For N02. 8.51 and 15.51

  For 03. 11.51 and 231
                                                                                                                                             Remarks
                                                      Haxliwm absolute percentage
                                                      deviations for the increase
                                                      and decrease, respectively.
                                                        For NO. 17land 401

                                                        For N02. 741 and 551

                                                        For 03. 91 and 1)1

                                                      The effects of Increaslno
                                                      and decreasing emissions
                                                      rates are alMOst Identical;
                                                      peak basin-wide average per-
                                                      centage changes In CO and
                                                      NO? nere about the sane
                                                      (6-81)
                                                                                        The base case value MS S
                                                               The buildup of the Mixing depth
                                                               variation effect is  time
                                                               dependent.

                                                               Decreasing the Mixing depth has
                                                               a greater effect on  the ground-
                                                               level concentrations than 1n-
                                                               creasinq It; this result Is More
                                                               pronounced for reactive pollu-
                                                               tants.

                                                               The effect of changing the Mix-
                                                               ing depth 1s not uniform over
                                                               the Modeling region; It varies
                                                               fron place to place.

                                                               The effect on ground-level con-
                                                               centrations of changing the
                                                               Mixing depth Is roughly the sane
                                                               as that of changing  the wind
                                                               speed, as would be expected from
                                                               a dimensional analysis.

                                                               The effects of varying the
                                                               radiation Intensity  are tine
                                                               dependent.

                                                               The effect of changing light
                                                               intensity Is as significant
                                                               as that of changing  wind speed.
                                                               The study results are summar-
                                                               ized by the foilowlno ranking
                                                               of the relative iMportance
                                                               of the Input parameters (A •
                                                               Most important and 0 - feast
                                                               important):

-------
                                                     Table  ,V-t2 (Continued)
   Study
  Model Version
  and Attributes
Sensitivity Analysis
     Variations
Influence on ttedel Predictions
                                                                                                                                  Remarks
Reynolds et il. (1976)
SAI photochemical Model:
1973 version [see Liu
et •!. (1976i)]
Anderson et •!.  (1977)
SAI photochemical  mdel:
1977 version

  31 step carbon bond
  chemistry

  3-0 Mind field

  Lower microscale
  layer

  Lanb and Liu c!1ffu-
  slvlty algorlttais

  30 K 30 i 7 grid

  SHASTA numerical
  method

  Surface rcMovat

  Three-dimension*1
  Initial conditions
 Uniform wind velocities
 with height  were conpared
 with vertical  variation
 in horizontal  winds given
 by a power law formulation
 Hind speeds were reduced
 by 33X
                                                         Nixing depths were re-
                                                         duced by 33X
                                                         Hind speeds and Mixing
                                                         depths were both reduced
                                                         by 331
                                                         Emissions in suburban
                                                         areas surrounding Denver
                                                         were reduced ZSX with
                                                         weighted emissions in-
                                                         crease) in other areas
                                                         to make overall regional
                                                         missions equivalent to
                                                         those in the bast case
Parameter or
Variable
Hind speed
Horizontal
dlffusivlty
Vertical
dlffusivlty
CO
A
D

C

NO
A
0

C

°3
A
D

C

NO,
A
D

C

   The max (mum average percent-
   age deviations were:  28.5S
   for NO. 15X for NO?, 24S for
   CO. and 14X for 03

   The MxiNM average deviations
   in pphu were:  -0.35 for NO.
   -1.1 for NO,. -4 for CO. and
   -2 for 03  z

   The MaxiMM deviations In pph»
   were:  7.5 for NO, 15 for NO..
   30 for CO. and 26 for 0}    *

   Maximum predicted ozone In-
   creased bv 41; maximum area
   for which [03] i O.OB ppm
   Increased by 12S
                                      predicted ozone In-
                              creased by 161; maximum
                              area for which [Oj] i O.OBX ppm
                              increased by 71

                              Naxlmm predicted ozone in-
                              creased by 331; maximum area
                              for which [0,] > 0.08 PPM
                              increased byJ30I

                              No difference occurred in the
                              time, location, or magnitude
                              of Maximum ozone concentration;
                              differences among predicted
                              ozone concentrations In all
                              runs were not More than 0.010
                              PPM in at Most one or two grid
                              cells
                                                                                                                          Mixing depth    B •   B   B   B

                                                                                                                                         D   A   A   B
                                                                                              Radiation
                                                                                              intensity

                                                                                              Emissions
                                                                                              rate
                                                                                                                                         B
                                                                                        B   B
The effects of including wind
shear were similar to those
of increasing surface wind
velocities by roughly 2SS
because velocities within the
mixed layer are Increased be-
tween 20 and 70S of the surface
values as a result of shear
                                     A synergiSM exists between
                                     wind speed and Mixing depth
                                      In each scenario, no more than
                                      7S of the region-wide emissions
                                      were redistributed; changes of
                                      this size in the spatial dis-
                                      tribution of emissions has
                                      little effect on secondary pol-
                                      lutants such as ozone

-------
                                                   Table   V-2  (Continued)
     Study Group
Klllus (197?)
(private coMMjnicatlon)
   Model Version
   •IK! Attributes
SAI photochemical model:
1977 version (see
Anderson et •!. (1977)
for node! attributes]
  Sensitivity Analyses
	Variations	

{Missions in the Denver
Metropolitan area were
reduced 17.51 with a pro-
portional increase in
suburban areas to Mke
regional tMisslon* levels
equivalent to those In
the base case

Grid spatial resolution
MBS  relaxed from 1  x  1 mile
 to 2 x 2 Miles
Influence on Model Predictions

No difference occured in the
liMe. location, or Magnitude
of MaxtMM ozone concentration;
differences aMong predicted
ozone concentrations in  all
runs were not More than  0.010
PPM in at Most one or two grid
cells

The coarser grid resolution led
to no noticeable change  in the
tine to peak NO. NOj. and Oj
concentrations; the Magnitude
of peak concentrations MBS re-
duced for NO (691). N02 (21t),
and 03 (131)
                                                                                                            irks
By the tiMe ozone fonts.  Its
precursors have been distri-
buted over a Much greater area
than their source regions;
accordingly, the Influence of
Increased grid size on ozone
predictions should'be less than
that for priMary pollutants
such as NO
Anderson (1977)
(private cnneunication)
SAI photochemical rndel:
1977 version [see
Anderson et al. (1977)
for Model attributes]
 NO emissions  fro* a point
 source Mere Increased by
 20f (note that the source
 contributes roughly U of
 the Denver regional NO.
 burden)
The MaxiMUM Impact of increased
source missions anywhere in
the Modeling region was an in-
crease in hourly averaged NO
and NOj concentrations (12 and
5 ppb, respectively) and a
decrease In 03 (-4 ppb)
The effect was decidedly local
and did not Influence  peak
oxldant concentrations

-------
           CHAPTER VI.  MODEL APPLICATION FOR AIR QUALITY PLANNING
     Model application refers here to the use of an air quality model to
evaluate a specific air quality problem and to formulate control strategies
which, when Implemented, will ameliorate 1f not eliminate the problem.  From
a regulatory viewpoint, the question the model application phase attempts to
address 1s:  how much control 1s needed to reduce hydrocarbon emissions to a
level at which the ozone standard 1s attained?  An ancillary question is:
what effect does nitrogen oxide control have on ozone levels?  The air quality
planner would also Hke to know, for a given overall regional emission reduc-
tion, what source categories are most effective to control and to what level.
All of these questions can be addressed in an Airshed Model Application.
     Before formulating control strategies, the air quality planner will
want to know what the situation 1s likely to be in future years, 1n terms of
ozone levels, if no additional controls are adopted.  This 1s determined by
applying the Airshed Model using the baseline projection emission inventories
described in Chapter III.  Data Needs.  A preliminary estimate of the level
of additional control needed can then be made.
     Two approaches have been used to formulate candidate control strategies,
one more rigorous than the other.  The simpler approach is to use the
                                         i
Airshed Model in a so-called "sensitivity" mode whereby Incremental across-
the-board emission reductions are made and the resulting ozone concentrations
are simulated.  By employing this procedure, one attempts to "narrow in" on
the overall emission reduction needed and to subsequently select the control
measures necessary to provide that reduction.  The preferred procedure is to
                                    92

-------
simulate actual control measures in the context of an overall strategy.
By simulating alternate strategies, an optimum strategy may be Identified.
Whichever approach 1s taken, the final set of control measures should be tested
with the Airshed Model 1n order to demonstrate attainment.
A.   PREPARATION OF FUTURE YEAR EMISSIONS
     Whether a simulation of future air quality without additional controls
or whether a simulation of candidate control strategies 1s to be conducted,
one must first project the corresponding level of emissions resulting from
growth, controls, or both.  In the former Instance, the projected emission
Inventory 1s called a baseline projection while 1n the latter 1t 1s called
a strategy projection.  Normally one baseline projection 1s performed for
each projection year, usually 1987, and a later year (1995 for example).
However, some urban areas may want to examine the effects of alternative
levels of economic and population growth.  In such cases, multiple baseline
projection scenarios may be needed.
     Several strategy projections will also need to be performed for each
projection year.   These projections reflect varying combinations of
Individual control measures, such as RACT (reasonably available control
technology), LAER (lowest achievable emission rate), I/M  (Inspection and
maintenance), TCM's (transportation control measures), and possibly others.
The strategy projections also reflect which sources would be controlled
and when the controls would come Into effect.  If only final strategy
projections are to be performed, having been formulated through use of the
Airshed Model 1n a "sensitivity" mode, only one or two projections may need
i
to be made.  On the other hand, 1f a final control strategy  1s formulated
by simulating specific candidate strategies, a considerable  number of
alternate strategy projections will need to be made.
      It may also be desirable to examine a candidate  strategy  in the  base year,
                                     93

-------
     As discussed  1n Chapter  IV, Preparation of Model Input Data, the
 EMISSIONS and PTSOURCE files  contain hourly emissions of eight pollutants.
 Furthermore, these emissions  are 1n grldded form on the EMISSIONS file,
 there being no distinction between one source and the next.  Obviously,
 neither baseline nor strategy projections can be performed at this level.
 As discussed 1n Chapter  III,  Data Needs, stationary source projections
 are made on an annual basis by source category.  Mobile source projections
 are made for the entire  transportation network down to the link level 1n terms
 of average dally traffic and  peak hour traffic.  The emissions, whether sta-
 tionary or mobile, are recalculated and are disaggregated temporally.  Hydro-
 carbons and nitrogen oxides are apportioned to the five carbon bond categories
 and to nitric oxide or nitrogen dioxide, respectively.  Finally, the mobile
 source, point source, and area source emissions are assigned or allocated to
 the appropriate grid squares.  Essentially then, performing a projection of
 emissions for a future year 1s equivalent to regenerating the hourly grldded
 Inventory for model Input.
 B.   FUTURE YEAR MODEL SIMULATIONS
     Once the Airshed Model has been verified using one or more days of
 aerometric data and once the  baseline emission projections have been performed,
 one Is ready to simulate future ozone levels.  Preferably more than a single
 day has been verified.   If so, simulation of future year baseline emissions and
 subsequent formulation of control strategies should usually be based on the
 day having the highest ozone  concentration.  However, 1f the highest day
 occurred not as a result of urban generated ozone but rather as a result of
 transport, one will also want to formulate and evaluate control strategies on
a day for which urban generated ozone predominates.
                                    94

-------
     In addition to the projected emissions data,  the  starting  points  for
assembling model Inputs for future year simulations are the air quality
and meteorological data files for the day or days  selected for  control
strategy analysis.  All modifications made to  these files during the course
of the verification analysis should  be reflected 1n the Input files used
for future year simulations.  The meteorological files, namely  WIND, DIFFBREAK,
REGIONTOP, TEMPERATUR, METSCALARS, and TERRAIN, are used directly and  are  not
further modified In any way.  In  contrast, the air quality files, namely
AIRQUALITY, BOUNDARY, and TOPCONC, must be adjusted to correspond to antld-
          u
pated future year conditions, as  discussed below.

   1.  Baseline Simulations
       Simulating baseline conditions 1n a future  year corresponds  to  a scenario
1n which growth occurs but no controls other than  those having  previously
been adopted are considered.  Data preparation programs are  run using  the
projected baseline emissions to create the EMISSIONS and  PTSOURCE  files for
model Input.  (These files were described previously 1n Chapter IV, Preparation
of Model Input Data).  The files  containing air quality data may or may
not require adjustment.  Under a  baseline scenario, sources  outside the urban
area may have no controls beyond  those 1n place during the base year.   In this
case, an assumption could be made that transport  at the upwind edge of the
modeling region and aloft remains the same as  1n  the base year, thereby
requiring no adjustments to the BOUNDARY or TOPCONC  files.
     The AIRQUALITY file generally should be  adjusted, however.  This file con-
tains the Initial ambient concentrations for  each pollutant at the beginning of
the simulation.  Initial conditions  could be  assumed  to be proportional to the
        ".}-.'•' °
urban regional emissions.  Future year Initial conditions of hydrocarbons
could then be approximated by taking the ratio of the regional future year
                                   95

-------
 hydrocarbon emissions to the regional base year emissions and applying
 this factor to the  Initial conditions for all five carbon-bond categories.
 However, natural background levels should be subtracted out first so that
 only the anthropogenic portions of the Initial conditions are adjusted.
 In addition,  1f any pronounced temporal variation exists 1n the emissions
 ratio, this should  be taken Into account.  Nitric oxide and carbon monoxide
 Initial conditions  could be approximated similarly using NO  and CO emissions,
                                                           ^
 respectively.  Unfortunately* there 1s no such direct method for estimating
 future ozone  and nitrogen dioxide Initial conditions or those of other
 secondary photochemical pollutant species (PAN and BZA).  Since the accumu-
 lation of these pollutants 1n the atmosphere 1s generally limited by the
 amount of reactive  hydrocarbons emitted, use of a future year to base year
 hydrocarbon emission ratio could serve as a surrogate factor.
     Model results  may be presented as a printed map of ozone concentrations
 as shown 1n Figure  VI-1 or as ozone Isopleths as shown 1n Figure VI-2.  The
 Airshed Model Itself produces printed maps only; one number 1s plotted for
 each grid cell.  Maps may be generated for a selected hour at any vertical
 grid cell level, but normally only ground-level concentrations are of Interest
 when evaluating the need for controls.  Computer plotting routines external
 to the model are required 1n order to display ozone Isopleths.

     2.  Control Strategy Simulations
     As Indicated earlier, two different approaches to control strategy develop-
ment may be taken.  One 1s to use the Airshed Model 1n a "sensitivity" mode.
Across the board reductions 1n hydrocarbon and/or nitrogen oxide emissions
are made to varying levels.  The principal advantage of this method 1s that 1t
 Is relatively easy to do.  The principal disadvantage 1s that 1t Ignores the
capability of the model  to evaluate the effect of significant changes 1n the
                                   96

-------
   POLLUTANT: 03     DAY:  159  HOUR:  14-15

              AVERAGE  AMBIENT  CONCENTRATIONS 
-------
                                                              MIRTH
vo
00
                                                        •••••••••••••••••••••••••••••••tMi ••••••••••

                                                                                               30
                                                              SflUTH
                 FIGURE  VI-2. EXAMPLE OF .A PLOTTED CONTOUR MAP OF GROUND-LEVEL  OZONE CONCENTRATION
                               ESTIMATES.  (Tesche and Burton, 1978).

-------
locations of emission sources and 1n the reactivity of the  emissions.   The
final strategy selected on this basis may either be more  stringent  than neces-
sary or less stringent depending on the manner In which the overall  emission
reduction 1s actually distributed among sources, whether  elevated or ground-
level, more or less reactive, upwind or downwind.   The preferred method 1s
                                                                \
to project the effects of different combinations of control  measures on future
year emissions, on a source by source basis  as described  above  1n Section
A, Preparation of Future Year Emissions.  This method  allows one to use the
Airshed Model to Its full potential.
     Whichever approach 1s taken, 1t 1s helpful  to  have some notion before-
hand of the approximate level to which total regional  hydrocarbon emissions
must be limited 1n order to meet the ozone standard.  The Empirical Kinetic
Modeling Approach (EKMA) offers a simple technique  (EPA,  1977c).  Using the
EKMA standard curves, which show the relationship between ambient levels of
hydrocarbons and nitrogen oxides and the level of ozone under a set of standard
conditions, one may estimate an approximate  percent hydrocarbon reduction
needed to reach the ozone standard.  Two pieces of  Information are  required:
the ozone "design value" and the NMHC/NO  ratio. The  design value, as
                                        n
described 1n "Guideline for the Interpretation of Ozone A1r Quality Standards"
(EPA, 1979), 1s the ozone concentration which has an expected exceedance of
once per year.  The design value 1s determined statistically Irrespective of
the particular day selected for control strategy analysis.   However, the NMHC/NO
ratio should correspond to the mean of the urban (as opposed to the upwind
or downwind) air quality monitoring sites for the 6 to 9 a.m. period on the
day or days selected for control strategy analysis.  Once an approximate
percent hydrocarbon emission reduction for the base year 1s estimated  using
the EKMA curves, this may be translated Into a percent reduction for any
                                    99

-------
particular baseline projection year.  This 1s done by first computing the
reduced base year regional hydrocarbon emission total and dividing by the
projection year baseline emission total.  This of course 1s subtracted from
1.00 and then multiplied by 100 to obtain the percentage reduction.  This
provides a first estimate of the overall level of control that may be needed.
With this Information one can begin to Identify the control measures that may
be necessary and to formulate candidate control strategies.
     Boundary conditions for the modeling region are Important regardless of
the approach taken.  Equity requires that a downwind city should not have to
exercise additional controls merely because an upwind city falls to meet Its
responsibilities to attain the ozone standard.  Therefore, the air quality
planner will want to assume the upwind city 1s meeting the standard.  How
then can the boundary conditions be made to reflect this?  An Ideal approach
would be to use a regional photochemical model.  A regional model 1s now
under development by the Environmental Sciences Research Laboratory, U. S.
Environmental Protection Agency, Research Triangle Park, North Carolina, but
1s not yet available.  A simple technique 1s to assume that the change in
ozone transport aloft and along the upwind edge of the modeling region 1s
proportional to the change 1n ozone levels 1n the upwind city.  Stated
another way, the ratt& of the upwind city "design value" for ozone during
the base year to the ozone standard 1s the same as the ratio of the ozone
boundary value 1n the base year to the boundary value 1n a future year.  If
the maximum ozone level for the upwind city 1s known for the specific day
being modeled, that value should be used in place of the design value.
Transported precursors (hydrocarbons and nitrogen oxides) should probably not
be reduced unless it can be demonstrated the upwind city 1s the source of the
emissions.  Generally, the relatively low levels of transported precursors
                                   100

-------
normally observed are thought to have their origin 1n nonurban areas
Immediately upwind of a metropolitan area.   Once an overall  reduction 1n
transported ozone has been estimated, the BOUNDARY and TOPCONC files must
be recreated to reflect the change.
     If the sensitivity approach 1s  used to develop control  strategies,  the
emission reductions selected should  generally correspond  1n  an overall sense
to specific control measures.  For example, preliminary analyses  may Indicate
hydrocarbon reductions 1n the 25 to  50 percent range, relative to future year
baseline emissions are needed.  Rough estimates of the effect of  "RACT"  on
point sources might show that a 22 percent  reduction 1n total regional hydro-
carbons could be achieved while "LAER" might provide another five percent.
Application of a strong "I&M" program might reduce regional  emissions by four
percent and a combination of "TCMs"  another two percent.   In this example,
one might select overall reductions  of 22 percent (RACT)  33  percent (LAER/
I&M/TCMs) and 50 percent.  The gap between  30 and 50 might,  for Instance, be
ascribed to new source review Incorporating a stringent offset policy.
     Across-the-board emission reductions such as these are  readily accomplished
through the use of techniques provided 1n the data preparation programs.  The
EMISSION FACTORS packet 1s suitable  for this purpose.  Initial concentrations
contained on the AIRQUALITY file should also be reduced by the same factor as
the emission rates.  This is done by adjusting the station readings used as
input to the data preparation program; the  program is then reexecuted to
create a new AIRQUALITY file.
     Assuming the Airshed Model simulation  runs show that the ozone standard is
met and exceeded (i.e., "bracketed") within the range of hydrocarbon reductions
selected, one may Interpolate between runs  to estimate the overall percent
reduction required to Just reach the standard.  It 1s then necessary for the
air quality planner to select what combination of control measures applied to
                                  101

-------
which sources with what level of control are needed to attain the standard.
Alternate sets of measures should be tentatively selected because the key
step 1n the sensitivity approach 1s confirming one of these strategies by
means of actual Airshed Model simulations.  This 1s done by performing detailed
emission projections (as described above 1n Section A.  Preparation of Future
Year Emissions) followed by running of the Airshed Model.  The model should
be run for each of the high ozone days previously verified.  Hopefully, one
of the strategies chosen will show attainment on all test days.  If not, more
stringent measures are needed and the process 1s repeated.
                                                        i
     The alternate and preferred approach differs primarily In the manner
by which control strategies are formulated.  Rather than making area-wide
emission reductions, one performs detailed projections of candidate strategies
composed of various control measures individually or in combination.  Com-
bining them 1n different proportions to different source types, one may arrive
at any number of candidate strategies.  For example, "RACT" reductions might
be applied to refineries and surface coating operations but not to service
stations in one strategy while another strategy might specify "LAER" reduc-
tions for all three.  These candidate strategies are then simulated using
the Airshed Model.  The results are then evaluated 1n the context of the
strategy's potential for attaining the ozone standard.
     Once the candidate strategies have been simulated and the results
analyzed, the air quality planner is in a good position to select those
control measures and the source types to which each should apply which 1n
combination will attain the standard.  If the candidate strategies have been
carefully selected in relationship to one another, a final strategy may
be arrived at readily with a minimum of trial and error while assuring,
when Rested, attainment will indeed be shown.
                                  102

-------
     Regardless of the approach taken, once a "final"  strategy has  been
selected, 1t should be tested to confirm that 1t  will  provide for attain-
ment of the ozone standard.  A detailed projection  of  the effect of the
strategy on emissions as previously described should be  performed.   Air-
shed Model simulations are then conducted on several previously verified
high ozone days, not just the one or two days adopted- for strategy  formu-
lation, 1n order to test the strategy under several meteorological  and air
quality scenarios.  This 1s Important because differing  levels of ozone
transport from upwind urban areas and differing levels of ozone carry over
from a previous day, as well as differing meteorological  conditions during
the days modeled, will all Influence the effectiveness of the particular control
strategy selected.
     Airshed Model results from control strategy  simulations may be presented
1n various ways.  In addition to the printed maps and  isopleth plots men-
tioned previously, graphical techniques for comparing  strategies directly
are useful.  Figure VI-3 shows a "deficit enhancement" plot on which Isopleths
of the difference 1n ozone concentrations between two  control strategies or
between a control strategy and a baseline projection  are shown.  Having such
a plot, the magnitude and spatial distribution of the  relative effect of a
particular control strategy may be seen at a glance.   Figure VI-4 Illustrates
the temporal effects of a sensitivity run on maximum  ozone levels over the
course of simulations on two different days.  In  this  example, a 30 percent
reduction 1n both NOX and hydrocarbon emissions reduced the peak ozone con-
centration about 15 percent.  However, the timing of the peak concentration
1s unchanged.
     The ozone Isopleths resulting from different control strategies may be
analyzed further by determining at hourly Intervals the area over which
                                  103

-------
                                                   NBRTH
                                                         20
30
                                                   SBUTH
FIGURE V1-3. DEFICIT ENHANCEMENT PLOT  ILLUSTRATING THE  PREDICTED  EFFECT OF  A CONTROL  STRATEGY
             (Tesche and Burton, 1978).

-------
i"
   *
s
5  is

I.
              II    I
                  O—-O  »ASE CASE

                  O——a  30 KRCENT KOUCTION
                         IN ALL EMISSIONS
                                                                   3 AUGUST 1976 METEOROLOGY
I    I    I    I
I"
   10
i
   0
                  28 JULY 1976 METEOROLOGY
         !   f
B
1   !    i
         \
5
5
                                   Ttat of toy of NMrly lotwttl
     FIGURE VI-4.  TEMPORAL  EFFECTIVENESS OF A SENSITIVITY RUN ON PEAK OZONE  CONCENTRATIONS
                  (Anderson, et al., 1977).

-------
estimated ozone levels exceed various threshold values.  The results may then
be summarized graphically as shown 1n Figure VI-5.  Differences between two
otherwise similar strategies may be readily shown 1n this way.  For example,
one strategy might achieve greater reductions over a greater area than another
                               .>
even though both provide nearly the same reduction 1n peak ozone levels.
(Note that 1n this example, controls on NO  emissions from power plants cause
                                          A
higher urban ozone levels, a phenomenon which follows from the scavenging effect
nitric oxide has on ozone.)
     The results from Airshed Model simulations may also be translated Into
population exposures as shown on Figure VI-6.  The model generates hourly
ozone concentrations 1n each ground-level grid cell.  These can be used In
conjunction with grldded population values (previously estimated during the
development of area source grid allocation factors) to compute the number of
person-hours above various ozone threshold values.  The population exposure
curves are then drawn by plotting these values for each strategy or each year
against the corresponding ozone concentration.
C.  INTERPRETATION OF MODEL RESULTS
     Two Issues arise in regard to the application of the Airshed Model and
other photochemical models to control strategy analysis.  The first Issue
relates to model performance.  No model can be expected to give results which
are totally free of bias.  Even on days for which a successful verification
has been achieved, some residual bias will undoubtedly remain.  The second
Issue 1s the comparability of the model results to the ozone standard.  The
standard 1s expressed 1n terms of an expected exceedance while the model
results represent concentrations for a particular high ozone day on which
the model has been run.  Both of these Issues should be addressed when

                                   106

-------
§
2
•M
§
U
§
o
01
i
O tf)
•M 10
     15


     10


      5


      0


     40
^^ ^0  *
2-3  .
o
T-
*
o
°=*  20
01 -o
     10
 0)
 3
                                                                  pphm
10    11
TT    17
12
TI
13
T?
14
                                             15
                                                   16
                                                   T7
                         Time of Day (PST)
                                                                  pphm
                                                                   BASE CASE
                                                                   POWER PLANT
                                                                   MOBILE SOURCE
                                                                  pphm
     FIGURE VI-5. SIMULATED AREAL  INFLUENCE OF TWO EMISSION CONTROL STRATEGIES
                  ON 03 CONCENTRATIONS ABOVE THREE CONCENTRATION LEVELS
                  (Tesche and Burton, 1978).
                                107

-------
o
00
I
W




I
                                                              1976


                                                              1985
                                              8   10    12    14   16   18   20    22   24   26   28


                                                     Oiont Conctntratlon (ppiw)
                FIGURE VI-6.  EXAMPLE OF THE  RESULTS OF A  POPULATION EXPOSURE ANALYSIS USING  THE AIRSHED MODEL

                              (Anderson, et al., 1977).

-------
applying the model to control  strategy analysis  to  avoid  either over-
stating or understating the level  of control  required  to  meet  the  ozone
standard.
     The best way to eliminate model  bias  1s  to  adjust the model Itself
to remove the bias at Us source.   However, such "fine tuning" of  the model
could be quite time consuming  and  may be  Impractical under the time and
resource constraints Imposed upon  the modeling effort. Calibration offers
an alternative to reformulating the model.  Calibration attempts to adjust
the model ozone estimates themselves to better correspond with ozone
observations.  Calibration does not necessarily  Improve the  accuracy of the
model; rather 1t simply ttrles to  compensate  for any bias present  1n the
results.  Calibration 1s not a substitute  for good  model  performance.
     The second model Interpretation Issue arises when the prototype day has
a different maximum ozone concentration that  does the  design day.  The
prototype day 1s the particular day on which  the model 1s run.  The design
day 1s a hypothetical day on which the ozone  design value occurs.  The  design
value 1s the ozone concentration which has an expected exceedance  of once
per year.  Methods for determining the design value are discussed  1n
"Guideline for the Interpretation  of Ozone A1r  Quality Standards"  (EPA,
1979).
     There are no assurances that  a day with  an ozone  concentration  equal
to the design value will occur during the three-month  aerometric data
collection study.  Typically,  the  day or  days selected for  modeling  will
have a maximum observed ozone  concentration  less than  the design value.
Therefore, the ability to demonstrate on  a prototype  day that the ozone
standard 1s not exceeded under a given emission control strategy does
not necessarily assure that the ozone standard  will not be exceeded on
                                   109

-------
the design day.  The only entirely satisfactory way to address the design
day Issue 1s to actually evaluate a given strategy using the Airshed Model
on the design day.  However, given that the prototype day does not corres-
pond to the design day, an objective technique 1s needed to relate model
estimates on the prototype day to the design day.  The question becomes, how
can the model estimates be adjusted so they can be used for comparison with
the ozone standard?  An attempt at answering this question has been made by
the Association of Bay Area Governments (ABAG, 1979).
                                   110

-------
                CHAPTER VII.   RESOURCE  REQUIREMENTS  FOR
                         AN AIRSHED MODEL STUDY
     Use of the Airshed Model  1n  an urban area for control  strategy  planning
1s an expensive undertaking.   As  described 1n earlier chapters, a  large  data
base needs to be amassed first.   Emissions data must be collected, compiled,
reviewed, and analyzed.  Field studies  to gather the necessary meteorological
and air quality data must be  planned and  executed.   The resulting  data must
be compiled, reviewed, validated, and analyzed.  Each of  these two data  col-
lection activities Involve long lead times and are generally done  largely
                                                i
under contract.  Modeling Itself  requires specialized skills and experience
not commonly found 1n local pollution control agencies and  1s therefore
normally a contract activity.  Modeling verification may  be a lengthy process
but one which must be accomplished before control strategy  evaluation and
testing can begin.  Execution of  the Airshed Model requires the  use  of a
large high speed computing facility. Running times  are long and computer
costs are substantial.
A.   MODELING PERSONNEL
     While the number of persons  needed to perform the modeling  verification
and control strategy analysis phases of the overall  study is not necessarily
large, a diversity of knowledge and experience  1s  required.  The various
disciplines needed Include:
     - engineering
     • computer science
     - meteorology
     - chemistry
     - statistics
                                   111

-------
 Because personnel normally follow disciplinary lines, the modeling team
 should Include at a minimum an air pollution engineer, a systems analyst
 or programmer, an air pollution meteorologist, an atmospheric chemist,
 and a statistician.  The engineer and computer specialist are often Involved
 1n most aspects of the modeling.  The extent of involvement of the meteor-
 ologist, chemist, and statistician depends on their individual background
 and experience 1n modeling.  Their abilities are essential 1n certain key
 phases of the study, yet their special expertise is not necessarily required
 throughout the entire study.  For example, the skills of the meteorologist,
 chemist, and statistician are not normally needed during the control strategy
 evaluation phase.  However, any one of the team participants, regardless of
 their particular disciplines, may take on the lead role in the study depend-
 ing on his or her experience in photochemical modeling.
     The air pollution engineer should have experience in applying models and
 should be familiar with emission inventory practices and techniques.  During
 the control strategy evaluation phase, he has primary responsibility for pre-
 paring strategy emission projections, for preparing the model Input data, and
 for evaluating the model results.  During the model verification phase, he 1s
 involved 1n evaluating verification problems and conducting sensitivity
 analyses, particularly as related to emissions data.
     The air pollution meteorologist should have a background 1n boundary
 layer meteorology and modeling.  It 1s his responsibility to prepare the
meteorological  data for model Input.  This Includes specifying the wind field,
mixing heights, and stability data for the modeling region.  He 1s heavily
 Involved in the verification phase in evaluating the effect of meteorological
 variables on model  performance.  An 1n-depth analysis of the meteorological

                                   112

-------
data base and the use of several  data  preparation  schemes may  be  needed
before satisfactory results are achieved.
     The atmospheric chemist should have a  background  1n ambient  monitoring
and the chemistry of ozone formation.   He should be  especially familiar
with the carbon-bond mechanism and how 1t treats various organic  compounds.
The atmospheric chemist 1s responsible generally for assisting 1n the
assessment of the reasonableness  of the emission Inventory  and specifically
for reviewing how the organic emissions from various source types have been
assigned to the five carbon-bond  categories.   He 1s  also responsible for
preparing the ambient nonmethane  hydrocarbon data.  The chemist plays a  key
role 1n evaluating air quality data aloft and upwind 1n order  to  prepare
model boundary conditions.  He also assists 1n the preparation of Initial
conditions.  The atmospheric chemist Joins  1n the  verification effort  to
determine whether the air quality data 1s 1n need  of re-examination and
whether the model chemical kinetics are performing properly.
     The computer specialist 1s the backbone of the modeling team.  It 1s
his Job to carry out many of the  data  manipulations necessary to  create the
files requested by the other project team members.  Frequently, this may
Involve modifications to the existing  data  preparation programs or entire
new programs may need to be written.  Most  of the  data manipulations neces-
sary for performing strategy emission  projections, followed by the necessary
splitting, grlddlng, and recomblnlng,  must  be done using computer processing.
Obviously, the computer specialist must be  thoroughly familiar with the
computer system on which the model 1s  to  be exercised.  He must be know-
ledgeable concerning system design, data  storage,  tape handling  procedures,
and structured programming.  The computer specialist must, of course, know
                                  113

-------
how to operate the large Airshed Model system of programs and files.
Finally, he should have a working knowledge of computer plotting soft-
ware so that model outputs (and Inputs) can be examined and evaluated
graphically.
     The statistician 1s primarily Involved in evaluating model perform-
ance.  It 1s he who 1s responsible for Identifying the performance measures
to be used, for applying, and for Interpreting such measures.  He assists
1n examining the model results to isolate probable causes of verification
problems.  The statistician works with the computer specialist in manipu-
lating the data, either numerically or graphically.  He may also be called
upon at the beginning of the modeling study to assist in examining and
analyzing the data base, particularly the aerometric data.
B.   COMPUTER FACILITY
     The Airshed Model computer program 1s written entirely 1n FORTRAN.
                                                                        i
The executable module requires 61K 36-BIT words of memory for operation on
a UNIVAC 1144.  On the same computer, running time for the model 1s approxi-
mately 15 minutes per hour of simulation for a modeling region of 300 grid
squares and four vertical layers, I.e., 1200 grid cells altogether.  Since
running time depends primarily on the size of the modeling region, more
grid cells would require proportionally more time.  However, this time can
also vary somewhat due to the shorter time steps necessary for Increasing
wind speeds.  If the user-defined modeling region is too large to fit into
the available core of the host computer, a segmentation process 1s avail-
able which divides the region Into one or more rectangular segments.  Using
segmentation, the data arrays for all segments are maintained in secondary
storage while the arrays for the segment being processed reside 1n core.

                                 114

-------
The times and core requirements given above reflect a modeling region
divided Into two segments.
     Initially the model requires the operation of eleven preprocessor
programs.  Each of these programs performs a similar function 1n that each
prepares and formats the various types of data  required  for Input to the
model.  Each program requires approximately 60K 36-BIT words of memory for
operation on a UNIVAC 1144.  Also, each program requires an average of
approximately two minutes of running time.  The files produced by these
programs can be written on either magnetic tape or disc.  Eleven output files
are produced for Input to the model.  In addition  to these eleven, the seg-
mented modeling region described above requires the use  of seventeen external
mass-storage files requiring an overall total of approximately 725,000
36-BIT words.

C.   PROJECT TIMETABLE
     A project the size of a photochemical modeling study using the Airshed
Model poses considerable uncertainty not only 1n regard  to Its cost, but also
in regard to the time required for Its completion.  Figure VII-1 represents
the kind of study timetable one may expect.  Overall, a  study using the
Airshed Model may be expected to take three years  or more.
     Once the local pollution control agency has made the decision to perform
an Airshed modeling study, Initial planning begins,  this may take about
six months.  An analysis of the existing emissions data base and the
historical meteorological and air quality data must be carried out.  With
this Information, decisions can be made on the dimensions of the modeling
region and on the exact data needs for the study.   The adequacy of the
existing monitoring network and the need for additional meteorological  and
air quality monitoring sites and Instrumentation must be assessed.  The
                                   115

-------
YEARS 0 1 2
i i
MONTHS
TASKS '
INITIAL PLANNING
DATA COLLECTION
Contract Planning
AQ/MET Field Study
AQ/MET Data Compilation
Emission Inventory Data Collection
Emission Inventory Compilation
MODELING
•> Contract Planning
Performance Evaluation
Strategy Evaluation
1 1 1 1 1 1 1 1
1 3 6 9 12 15 18 21 24
i



*
i i
i i
•
I1
3

27 30 33 36
o , ,,.n|
I 	 	 h.

Figure VI1-1.  A Photochemical Modeling Study timetable.

-------
requirements for upper air meteorological  and air quality data must be estab-
lished.  The extent to which the existing  stationary source emission Inventory
meets the needs of the study must be  assessed and additional data needs must
be Identified.  The level  of effort required to take the existing transportation
data base and use 1t for generating the mobile source emission Inventory must
be estimated.  The assistance of a contractor familiar with the Airshed Model
and Us requirements 1s a  valuable part of the Initial planning effort.  Toward
the end of the Initial planning  phase, 1t  should be possible to make reasonable
estimates of the resources required to perform the study.
     Soon after the Initial planning, the  local agency will want to decide  on
the extent of contractor Involvement. This depends on the In-house man-hours
available to devote to the project and the availability of outside funding.
Contractors are oftentimes heavily Involved 1n the data collection effort.
Depending on agency contracting  procedures, contract solicitation, selection,
and award can require seven months of elapsed time, if all goes well.
     Once one or more contractors have been selected to operate supplementary
surface monitoring sites,  collect aircraft and upper air meteorological  data,
and analyze hydrocarbon grab samples, the  field study can begin.  For a  ,
three month data-collection period, one should probably count  on at  least
two months of preparation  time.   New  monitoring stations and associated
Instrumentation must be acquired, Installed, and readied.  Aircraft  must be
equipped, the flight crew assembled,  and flight plans drawn  up.  Radiosondes,
plbals, and associated tracking  equipment  must be obtained and release  sites
prepared.  Existing surface monitoring stations may need to  be reequlpped
or overhauled.  With the onset of the "oxidant season," data collection  begins.
After it 1s collected, the data  must  be validated, archived,  reviewed,  and
analyzed.  Before the data is ready for modeling, another  seven months  will
likely have passed.
                                   117

-------
     Concurrent with the aerometrlc data collection effort 1s the emissions
data collection effort.  Agency stationary source files must be reviewed
and updated as necessary.  Area source emissions must be determined for each
county 1n the modeling region and allocation factors must be developed to
allocate them to the grid system.  Temporal profiles must be developed for
each source as do profiles for apportioning hydrocarbon and nitrogen oxide
emissions to the five carbon-bond categories and to NO and N02-  Diurnal traffic
distribution data must be developed, VMT estimates made, and speeds computed
for the transportation Inventory.  Vehicle operating characteristics must be
estimated.  Approximately six months may be required for collecting these
various data.  Another six months may be needed to ready network emissions
models, to prepare other computer software, and then to generate the hourly
gridded emission Inventories for model Input.
     Well before the emissions and aerometrlc data bases are expected to be
ready for modeling, the agency should obtain the services of a modeling con-
tractor.  As for the data collection effort, seven months may be required to
consummate a contract.
     Model performance evaluation begins with an analysis of the data base.
The meteorological and air quality data must then be prepared for model Input.
Once the model 1s run, the results must be evaluated and any problems cor-
rected.  This requires about nine months.  The formulation of candidate
control strategies and the selection and testing of a final strategy are
open-ended activities which may last as long or longer than the model per-
formance evaluation.

D.   OVERALL PROJECT COSTS
     Table VII-1  shows a range of overall project costs, exclusive of com-
puter costs, broken down by major activities.  All cost Hems are assumed to be
                                     118

-------
                Table VII-1.   Photochemical  Modeling Study Costs
Activity
PLANNING
DATA COLLECTION
  Surface AQ/MET Data
  Upper A1r MET Data
  Aircraft AQ Data
  Organic Species AQ Data
  Emissions Data
  Subtotal
MODELING
  Performance Evaluation
  Strategy Evaluation
  Subtotal
TOTAL
       Cost1
Low  ($1000)

 20

250
 30

 20
150
450
                              Computer Time*
100
 80
180
650
H19h
40
550
100
150
40
850
1690
180
no
290
2020
Low (Hours) ^*
_
1 3
1 1
1
-
15_ 30_
17 35
7 12
JL li
1_5 28
32 63
1
 Contract costs, excluding computer costs, 1979 dollars.
 Tor a CDC 7600 machine.
                                     119

-------
 done outside the  local agencies, under contract.  Costs for managing the project,
 which may  require one or two man-years 1n and of Itself, are assumed to be
 absorbed by the agency's own manpower resources and not require external fund-
 Ing.  Any  tasks performed by the agency In-house will reduce project costs pro-
 portionally.  Also shown 1n Table VII-1 are approximate computer time requirements
 for a machine such as the CDC 7600.  Based on these estimates, total noncomputer
 costs for  a photochemical modeling study may range from a low of 600 thousand
 dollars to a high of 1.9 million.  Between 75 and 85 percent of this 1s for
 data collection while the remainder 1s for modeling and for contractor assistance
 in the initial planning of the project.  In addition, at a unit cost for com-
 puter time of two thousand dollars per hour on a machine such as the CDC 7600,
 total computer costs may range from 65 to 125 thousand dollars.
     Previous experience in planning photochemical modeling and aerometric and
 emissions  data collection studies shows that Initial planning costs may range
 from 20 to 40 thousand dollars.  This 1s the cost at the start of the project
 for contractor assistance 1n planning the overall study and of determining
 what the particular data needs are.
     Collection of surface air quality and meteorological data 1s usually the
                                                                       <*
 largest single cost Item in a photochemical modeling study.  As indicated in
 Table VII-1, costs may vary from 250 to 550 thousand dollars or more.  The
 cost depends on the number of new monitoring sites required and on any
 additional  monitoring equipment needed at existing state and local agency
 sites.  The cost figures shown here assume that all new sites are Installed
 and operated under contract.  No costs are Included for operating existing
monitors since these costs are already being Incurred by state and local
agencies.   In Philadelphia, (PA-NJ-DL-AQCR) for example, where an Airshed
Model  application 1s underway, operation of five new sites for three months
cost 250 thousand dollars.   This Includes data collection, compilation,
                                   120

-------
validation, review, and analysis  but  does  not  Include the cost of  Instruments,
which were provided to the contractor.   These  five sites were supplemented
by 17 state and local  sites.   Some  computer time  1s also required  to compile
and analyze the entire surface aerometHc  data base.
     Other cities where photochemical modeling studies are being planned are
New York, Boston, Baltimore,  and  Washington, D.C.  Estimated costs for
surface aerometrlc data range from  550  thousand dollars 1n New York to  350
thousand 1n Baltimore.  In Baltimore, four new sites are proposed  to sup-
plement seven existing sites  which  are  being upgraded while in New York
(NY-NJ-CN AQCR) the number of new and existing sites are 10 and 20 respec-
tively.
     Upper air meteorological data  consists of rawinsonde, plbal,  and
acoustic sounder measurements. Among cities now  being studied, cost estimates
range from 30 to 100 thousand dollars.   Equipment and operating costs  are
included.  The high estimate  of 100 thousand dollars is for Boston where no
sites now exist and a radiosonde  and  two plbal sites are planned.  At  the  low
end, New York, a radiosonde site  1s available; two additional  plbal  sites  are
planned.  In Philadelphia, 50 thousand  dollars was expended for a  single
radiosonde site at which hourly plbal  releases were also made.
     Measurements of air quality  data aloft using Instrumented aircraft
may or may not be taken as discussed  in Chapter  III.  Data  Needs.   It  has
been estimated to cost 150 thousand dollars for  15  in-flight  days.  This
also includes the cost of standby flight crews ready  to  fly during a typical
three month summer study.  However, these  are  only operating costs and do not
reflect the cost of equipping an  airplane  with instruments.
                                  121

-------
     A cost of 30 thousand dollars has been estimated for collecting 200
grab samples and conducting a complete organic species analysis using gas
chromotography.  The cost for a given study area depends on the number of
samples analyzed.
     The cost of developing an hourly gridded emissions Inventory for model
Input 1s not so much dependent on city size, either 1n terms of geographic
area or population, but rather on the condition of the existing data base.
Contract costs may range from a low of 150 thousand dollars to a high of 850
thousand dollars for an Inventory of mobile and stationary sources.  In
Baltimore, where the stationary source Inventory 1s relatively complete and
current, an estimated 70 thousand dollars 1s needed to derive the necessary
spatial and temporal resolution, to develop VOC and NO  pollutant profiles,
                                                      A
and to perform two baseline projections.  Another 95 thousand dollars 1s
needed to generate the mobile source emission Inventory, making the total
Inventory costs for Baltimore approximately 170 thousand dollars.  In compar-
ison, 1n the Metropolitan New York (NY-NJ-CN) AQCR, over half of an estimated
850 thousand dollar total cost Is needed just to collect the data necessary
to develop a basic annual Inventory of point and area sources.  Only 125
thousand of the total 1s needed for mobile sources.  In the metropolitan
Washington area, a total cost of 250 thousand 1s estimated of which 140
thousand 1s for mobile sources and 110 for stationary sources.  These costs
all Include the cost of computer processing.  In Baltimore, 28 thousand
dollars has been budgeted for computer processing to generate mobile source
emissions for model  Input from the transportation Inventory.  In Tulsa, 19
thousand dollars was expended for stationary source emissions computer
processing.
                                  122

-------
     In addition to the basic work of collecting  and  assembling  the emis-
sions data and manipulating 1t for model  Input, 1t  1s necessary  to oversee
closely the Inventory effort and to thoroughly review the  emissions data.
This quality assurance function, 1f done  under contract, may  cost between
25 and 50 thousand dollars.  However, these  costs are not  Included 1n
Table VII-1 where 1t has been assumed the local agency will perform this
function without external funding.
     The cost for model verification and  control  strategies depends on  the
number of Airshed Model runs and the size of the  modeling  region (and there-
fore the size of the data base).  The same 1s true  of the  estimated computer
time.  Modeling costs range from 180 to 290  thousand  dollars. Model verifi-
cation Includes the cost of analyzing the data and  preparing  1t  for model  Input
as well as the simulations themselves, exclusive  of computer  costs.  Also  shown
1n Table VII-1 are estimated computer time requirements.   At  least two  trial
simulations are assumed necessary to achieve satisfactory  model  performance
on any given day; however, more may be required.  Model  verification  on three
or four different days 1s also assumed.   Control  strategy  analysis  Includes
the cost of performing strategy emission  projections  and of running  baseline
simulations and candidate control strategy simulations.  Anywhere from five
to 15 runs may be required depending on what control  measures are desired to
be evaluated.  In addition, final strategy confirmation  runs on three or
four days are also Included.  It should be realized that 1n running a model
as large as the Airshed Model, a significant amount of computer time can be
expended on wasted runs, runs for which no meaningful output results, due to
various pitfalls and errors.  The experience of one agency with a different
model was that for every successful run,  an  average of 2.5 runs had to be
made altogether (ABAC, 1979).
                                     123

-------
                             REFERENCES


ABAC (1979).  "Application of Photochemical  Models 1n the Development of
    State Implementation Plans, Volume I:  The Use of Photochemical Models
    1n Urban Oxldant Studies," EPA-450/4-79-025, Association  of  Bay Area
    Governments, Berkeley, California.

Ames, J. et al. (1978).  "The User's Manual  for the SAI  Airshed  Model,"
    draft report for Environmental Sciences Research Laboratory, Office of
    Research and Development, U. S. Environmental Protection  Agency, Contract
    No. 68-02-2429, Systems Applications Inc., San Rafael, California.

Anderson, G. E., et al., (1977).  "A1r Quality 1n the Denver  Metropolitan
    Region:  1974-2000," EPA-908/1-77-002 Systems Applications Inc.,
    San Rafael, California.

BHggs, G. A. (1971).  "Plume Rise:  A Recent Critical Review,"  Nuclear
    Safety. Vol. 12, pp. 15-24.

Bucon, H. W., J. F. Macko, and H. 0. Taback (1978).  "Volatile Organic
    Compound (VOC) Species Data Manual," EPA-450/3-78-119, KVB Engineering,
    Inc., Tustln, California.

Demerjlan, K. L. (1976), "Photochemical A1r Quality Simulation Modeling:
    Current Status and Future Prospects," Paper 16-1, International Con-
    ference on Photochemical Oxldant Pollution and Its Control,  U. S.
    Environmental Protection Agency, Raleigh, North Carolina.

Duewer, W. H., M. C. McCracken, and J. J. Walton (1978).  "The Llvermore
    Regional A1r Quality Model:  II.  Verification and Sample Application
    1n the San Francisco Bay Area," J. Appl. Meteor., Vol. 17, pp. 273-311.

Eaton, W. C., et al., (1979).  "Study of the Nature of Ozone, Oxides of
    Nitrogen, and Nonmethane Hydrocarbons 1n Tulsa, Oklahoma, Volumes I, II
    and III," EPA-450/4-79-008a, b, c, Research Triangle Institute, Research
    Triangle Park, North Carolina.

EPA (1979).  "Guideline for the Interpretation of Ozone  A1r Quality Standards,"
    EPA-450/4-79-003, Office of A1r Quality Planning and Standards, Research
    Triangle Park, North Carolina.

EPA (1978a).  "User's Guide to MOBILE!:  Mobile Source Emissions Model,"
    EPA-400/9-78-007, Office of A1r, Noise and Radiation, U.  S.  Environ-
    mental Protection Agency, Washington, D.C.

EPA (1978b).  "Mobile Source Emission Factors, Final Document,"
    EPA-400/9-78-005, Office of Transportation and Land  Use Policy,
    U. S. Environmental  Protection Agency, Washington, D. C.


                                  124

-------
EPA (1977a).  "Compilation of A1r  Pollutant Emission Factors, Third
    Edition," AP-42 parts  A and  B, Office of Air Quality Planning and
    Standards, U.  S. Environmental Protection Agency, Research Triangle
    Park, North Carolina.

EPA (1977b).  "Procedures  for the  Preparation of Emission Inventories
    for Volatile Compounds, Volume I," EPA-450/2-77-028, Office of A1r
    Quality Planning and Standards, U.S. Environmental Protection Agency,
    Research Triangle Park, North  Carolina.

EPA (1977c).  "Uses, Limitations and Technical Basis of Procedures for
    Quantifying Relationships Between Photochemical Oxldants and Precursors,"
    EPA-450/2-77-021a, Office of A1r Quality Planning and Standards, U. S.
    Environmental  Protection Agency, Research Triangle Park, North Carolina.

EPA (1976).  "AEROS Manual  of Codes," EPA-450/2-76-005t Office of A1r
    Quality Planning and Standards, U. S. Environmental Protection Agency,
    Research Triangle Park, North  Carolina.

EPA (1975).  "Comprehensive Data Handling System, Emissions Inventory/Permits
    and Registration Subsystem (EIS/P&R) Program Documentation and Users
    Guide/ EPA-450/3-74-045a as updated, Office of A1r Quality Planning
    and Standards, Research Triangle Park, North Carolina

Engineering-Science (1979a).  "Emission Inventories for Tulsa, Oklahoma
    for SAI Model  Application,"  draft report for Office of Air Quality
    Planning and Standards, U.S. Environmental Protection Agency; Contract
    No. 68-02-2584, Engineering-Science, McLean, Virginia.

Engineering-Science (1976b).  "Emission Inventory Protocol Document  [for]
    Metropolitan Philadelphia Oxldant Study," draft report for Office of
    A1r Quality Planning and Standards, U. S. Environmental Protection
    Agency, Contract No. 68-02-2584, Engineering-Science, McLean, Virginia.

GHscom, R. W.  (1978). "Regional A1r Pollution Study:  Point and Area
    Source Organic Emission Inventory," EPA-600/4-78-028, Rockwell
    International  A1r Monitoring Center, Creve Coeur, Missouri.

Hayes, S. R. (1979).  "Performance Measures and Standards for A1r Quality
    Simulation Models," EPA-450/4-79-032, Systems Applications Incorporated,
    San Rafael, California.

HUlyer, M. J., S. D. Reynolds and P. M. Roth  (1979),  "Procedures for
    Evaluating the Performance of  A1r Quality Simulation Models,"
    EPA-450/4-79-033, Systems Applications, Incorporated, San Rafael,
    California.

Jerskey, T. N., and J. H.  Seinfeld (1976).  "Continued  Research  1n Meso-
    scale A1r Pollution Simulation Modeling—Volume  IV:  Examination of the
    Feasibility of Modeling Photochemical Aerosol Dynamics,"  EPA-600/4-76-016d,
    Systems Applications,  Incorporated, San Rafael,  California.


                                   125

-------
 Oerskey, T. N., et al., (1976).   "Continued  Research  1n Mesoscale A1r
     Pollution Simulation Modeling—Vol.  VII:  Mathematical Modeling of
     Urban Aerosol Dynamics," draft report for Environmental Sciences
     Research Laboratory, Office  of Research  and Development, U. S.
     Environmental Protection Agency,  Contract No. 68-02-2216, Systems
     Applications, Incorporated,  San Rafael,  California.

 Kill us, J. P., et al., (1977).   "Continued Research 1n Mesoscale A1r
     Pollution Simulation Modeling—Vol.  V:  Refinements 1n Numerical
     Analysis, Transport, Chemistry, and  Pollutant Removal," draft report
     for Environmental  Sciences Research  Laboratory, Office of Research and
     Development, U. S. Environmental  Protection Agency, Contract No. 68-02-2216,
     Systems Applications, Incorporated,  San  Rafael, California.

 Lamb, R. G. (1976).  "Continued  Research 1n  Mesoscale A1r Pollution Simu-
     lation Modeling—Volume III:   Modeling of Mlcroscale Phenomena,"
     EPA-600/4-76-016c, Systems Applications, Incorporated, San Rafael,
     California.

 Lamb, R. G., W. H. Chen, and J.  H.  Seinfeld  (1975).   "Numerico-Empirical
     Analysis of Atmospheric Diffusion Theories," J. Atmos. Sc1.. Vol. 32,
     pp. 1794-1807.

 Lamb, R. G., et al., (1977).  "Continued Research 1n Mesoscale A1r
     Pollution Simulation Modeling—Vol.  VI:  Further Studies 1n the Model-
     Ing of Mlcroscale  Phenomena,"  draft  report for Environmental Sciences
     Research Laboratory, Office of Research and Development, U. S.
     Environmental Protection  Agency, Contract No. 68-02-2216, Systems
     Applications, Incorporated, San Rafael, California.

 L1u, M. K. and J. H. Seinfeld (1975).  "On the Validity of Grid and
     Trajectory Models  of Urban A1r  Pollution," Atmos. Environ.. Vol. 9,
     pp. 555-574.

 Liu, M. K., D. C. Whitney and P. M. Roth  (1976).  "Effects of Atmospheric
     Parameters on the  Concentration of Photochemical Pollutants," J. Appl.
     Meteorol., Vol.  15,  pp. 829-835.

 L1u, M. K., et al.,  (1976a).  "Continued Research 1n Mesoscale A1r
     Pollution  Simulation Modeling—Volume  I:  Analysis of Model Validity
     and Sensitivity and  Assessment of Prior Evaluation Studies,"
     EPA-600/4-76-016a, Systems Applications, Incorporated, San Rafael,
     California.

 Ludwlg,  F.  L.  and E. Shelar  (1978).  "Site Selection for Monitoring
     Photochemical A1r Pollutants," EPA 450/3-78-013, SRI International,
     Menlo  Park, California.

MacCracken, M. C., et al.,  (1978).  "The Llvermore Regional  A1r Quality
    Model:  I, Concept and Development," J. App. Meteor.. Vol. 17,
    pp. 254-272.


                                     126

-------
MacCracken, M. C. and 6.  D.  Sauter,  eds.  (1975).   "Development of an A1r
    Pollution Model  for the  San Francisco Bay Area,"  "UCRL-51920 Vols.
    1 and 2, Lawrence Livermore Laboratory, University of California,
    Llvermore, California.

MacCracken, M. C. (1975).  "User's Guide  to the LIRAQ Model:  An A1r
    Pollution Model  for the  San Francisco Bay Area,"  UCRL-51983, Lawrence
    Llvermore Laboratory, University of California, Llvermore, California.

MIlHgan, R. J., et  al.,  (1979).   "Review of NO  Emission Factors for
    Stationary Combustion Sources,"  EPA 450/4-79-021, Acurex Corporation,
    Mountain View, California.

Pacific Environmental Services, Inc. (1979).   "Procedures for the Prepara-
    tion of Emission Inventories for Volatile Organic Compounds, Volume II
    Emission Inventory Requirements  for Photochemical A1r Quality Simu-
    lation Models,"  EPA-450/4-79-018, Pacific Environmental Services,
    Santa Monica, California.

Reynolds, S. D. (1977).  "The  Systems Applications, Incorporated Urban
    Airshed Model:  An Overview of Recent Developmental Work," Inter-
    national Conference on  Photochemical  Oxldant Pollution and Its  Control,
    EPA-600/3-77-001b, Systems  Applications, Incorporated, San Rafael,
    California.

Reynolds, S. D., et  al.,  (1979).  "Photochemical Modeling of Transportation
    Control Strategies.  Volume I.   Model Development, Performance
    Evaluation, and  Strategy Assessment," draft final report for Office
    of Research, FHWA, U.S. Department of Transportation, EF79-37, Systems
    Applications, Incorporated, San  Rafael, California.

         (1978).  "Application of the SAI Airshed  Model to the Evaluation
    of Alternative Population Growth Forecasts for the South Coast A1r
    Basin," EF78-124, Systems Applications, Incorporated, San Rafael,
    California.

         (1976).  "Continued Research 1n Mesoscale A1r Pollution Simulation
    Modeling:  Volume II:  Refinements 1n the Treatments of Chemistry,
    Meteorology, and Numerical Integration Procedures," EPA-600/4-76-016b,
    Systems Applications, Incorporated, San Rafael, California.

         (1974).  "Mathematical Modeling of Photochemical A1r Pollution—III,
    Evaluation of the Model," Atmos. Environ.. Vol. 8, pp. 563-596.

         (1973a).  "Further Development and Validation of a Simulation
    Model for Estimating Ground Level Concentrations of Photochemical
    Pollutants," Systems Applications, Incorporated, San Rafael, California.

     	(1973b).  "Mathematical Modeling of Photochemical A1r Pollution—I.
    Formulation of the Model," Atmos. Environ.. Vol. 7, pp. 1033-1061.
                                  127

-------
Reynolds, S. D.f T. W. Tesche, and L. E. Reid (1978).  "An Introduction to
    the SAI Airshed Model and Its Usage," draft report to Office of A1r
    Quality Planning and Standards, U. S. Environmental Protection Agency,
    EF78-53R3, Systems Applications, Inc., San Rafael, California.

Roth, P. M., et al.t (1974).  "Mathematical Modeling of Photochemical
    A1r Pollution--!!.  A Model and Inventory of Pollutant Emissions,"
    Atmos. Environ.. Vol. 8, pp. 97-130.

         (1971).  "Development of a Simulation Model for Estimating Ground
    Cevel Concentrations of Photochemical Pollutants," APTD-0914 (Also
    Appendices A-F, APTD-0908 thru 0913), Systems Applications, Incorporated,
    San Rafael, California.

Schere, K. L. and K. L. Demerjlan (1978).  "A Photochemical  Box Model  for
    Urban A1r Quality Simulation," Proceedings of the Fourth Joint Con-
    ference on Sensing of Environmental Pollutants, American Chemical
    Society, 6-11 November 1977, New Orleans, Louisiana.

Tesche, T. W. and C. S. Burton (1978).  "Simulated Impact of Alternative
    Emission Control Strategies on Photochemical 0x1dants 1n Los Angeles,"
    EF78-22R, Systems Applications, Inc., San Rafael, California.

THjonls, J. C. and K. W. Arledge (1976).  "Utility of Reactivity Criteria
    1n Organic Emission Control Strategies:  Application to  the Los Angeles
    Atmosphere," EPA-600/3-76-091, TRW Environmental Services, Redondo
    Beach, California.
                                    128

-------
                APPENDIX
TECHNICAL DESCRIPTION OF THE AIRSHED MODEL

-------
                            TABLE OF CONTENTS

A.  The Basic Equation                                         A-3
B.  The Numerical Solution Procedure                           A-5
C.  The Estimation of Turbulent D1ffus1v1t1es                  A-10
D.  The Treatment of Atmospheric Chemistry                     A-15
E.  The Treatment of Emissions                                 A-25
F.  The Treatment of Surface Uptake                            A-30
                                     A-l

-------
                TECHNICAL DESCRIPTION OF THE AIRSHED MODEL*
      The development of the Airshed Model  1s described 1n  numerous  reports
 which detail the derivation of the model  and the steps within  each  algorithm
 and  subroutine.  These reports were mentioned previously 1n  the  Foreword to
 this document.  Although quite useful  for developing an understanding of the
 model, their length makes their use difficult.   This appendix  provides a brief
 technical description of the model so  that users can become  familiar with the
 basic concepts and equations used 1n the  model  without having  to peruse all
 the  available detailed material.  Since this appendix 1s not Intended as a
 guide to the use of the model, none of the Input or  output functions of the
 model or data storage manipulations within the  model  are described.
 A.    THE BASIC EQUATION                    ,
      The basis for the model 1s the continuity  equation, which expresses the
 conservation of mass of each pollutant 1n  a turbulent fluid  1n which chemical
 reactions occur.  This equation can be written:
   a^     +^^ + i^+^^  j/   ifi\  + j/,  !!iU-L/*  !!i\
    3t          3X        3y        32   = 3X \*H  3X )    3y \*H 3y /   3Z \*V  3Z /
 I         I   L                     I  I                               	I
    Time            Advection                    Turbulent  Diffusion
 Dependence
                                        +    *i   •+    si         •        (1)
                                          L	I   L.   .   I
                                          Chemical   Sessions
                                          Reaction
where c^ represents the pollutant  concentration and  1s a function of space
(x.y.z) and time (t).   This  equation describing the  dynamic behavior of
reactive pollutants 1s  fully three-dimensional.   Further examination of the
 Adapted from Reynolds, Tesche, and Reid  (1978).
                                  A-3

-------
equation Indicates that the following physical and chemical processes are
considered 1n the Airshed Model:
     >  Pollutant advection.  The model can treat a fully three-
        dimensional wind field, where u, v, and w are the mean wind
        velocity components in the x, y, and z directions, respectively;
     >  Turbulent diffusion.  Pollutant transport resulting from the
        influence of atmospheric turbulence is treated through the
        use of the eddy diffusivity concept.  KH and Ky are the
        horizontal and vertical diffusivity coefficients, respectively.
     >  Chemical reaction.  The term R^ represents the net rate at
        which pollutant 1 is generated by chemical reactions.  The
        reaction rate is a function of pollutant concentration,
        temperature, and the intensity of ultraviolet radiation.
     >  Emissions.  The spatial and temporal distribution of the
        source emissions are treated 1n the term S^.  In the case
        of large point sources, the total effective plume rise is
        calculated to enable the appropriate spatial placement of the
        emissions aloft.
In addition, removal of pollutants by surface uptake processes is con-
sidered in the boundary conditions of Eq. (1).

     To derive Eq. (1), one must make three assumptions.  First, pollutant
transport effects due to molecular diffusion are small relative to those
attributable to turbulent diffusion.  Second, pollutant transport due to
turbulence can be adequately parameterized through the use of the eddy
diffusivity concept.  Third, turbulent concentration fluctuations have a
                                   A-4

-------
negligible Influence on reaction rates.   For a  more  thorough discussion of
the derivation of Eq. (1), the reader 1s  referred  to reports by  Reynolds
et al. (1973a, 1973b).
B.   THE NUMERICAL SOLUTION PROCEDURE
     Because Eq. (1) 1s nonlinear,  1t 1s  not possible to  obtain  an  analytical
solution.  Thus, one must utilize appropriate numerical techniques  to  find an
approximate solution.  To facilitate the  application of finite difference
methods, one first performs a change of variable to  normalize the vertical
dimension by the distance between the bottom and top of the region. This 1s
accomplished by defining a new Independent variable  p as  follows:
                         2 - Hb(x,y.t)
                  p  = Htu,y.tji  - Hbu,y,tj

where Hb and Ht are the elevations  of the bottom and top  of the  region,
respectively.  Upon performing this change of variable and neglecting
small cross-derivative turbulent diffusion terms*  Eq. (1) becomes
(Reynolds et al., 1973a, 1973b):
                                            * apW ap/   +
                                                                    (2)
where
            w . u
            w   u
ax     ax /  ~   \ ay     ay /  *   at     *
                                    A-5

-------
                       AH = Ht(x,y,t) - Hb(x,y,t)
     As indicated in Eq. (2), the Airshed Model requires values of u, v,  and
W in order to carry out pollutant advection calculations.  Actually, the
user must specify only the values of u and v 1n each grid cell, and the
Airshed Simulation Program computes appropriate values of W from the wind
continuity relationship and information pertaining to the depth of the
modeling region.  To calculate W given values of u and v in each grid cell,
one first writes the continuity relationship:
                                                                       <3'
Upon performing the same change of variable just described, one obtains:
                      3(UAH) , 3(VAH) t 3W- o
                         sx        ay     SP                             (4)
where
               --        /aHb     3AH\    /aHb +   3AH\
                                                                       (5)

The value of W used in Eq. (2) is then given by:
                              W = W  - P-^                            (6)

Equation (4) can be written in finite difference form and solved subject
to the constraint that fl = 0 at the ground.  Equation (6) is then used
to complete the specification of the values of W employed 1n the
numerical solution of Eq. (2).
                                    A-6

-------
     It should be noted that reliance on this technique for assuring a mass
consistent wind field should be undertaken with caution.  If the u-v wind
field within a layer 1s characterized by excessive divergence, as might
occur following use of a simple Interpolation scheme  applied to station
measurements, the divergence must be reduced by employing suitable tech-
niques for "smoothing" the wind field prior to the simulation program.
Otherwise, unreal1st1cally large vertical velocities may result.
     Special consideration must be given to the application of finite
difference methods to multidimensional problems.  The Airshed Model employs
the so-called method of fractional steps.  Applying this technique to the
solution of Eq. (2), one obtains the following governing equations for the
four-step numerical integration procedure:
    >   Step  I
               &*1>  *&«",>  •&(«,,•«§•)    •           (7)
    >   Step  II
               ^HC^^VAHC^^JVH^?-)    .          (8)
    >   Step  III
    >   Step IV
                                       RjAH                       (id
                                  A-7

-------
 If the modeling region  1s  divided  Into  two  layers, then there will be
 two sets  of Eqs.  (7)  through  (10),  differing  only in the definition of
 AH.  the  numerical  Integration  of  Eq.  (2) for one time interval 1s
 approximated in the method of fractional steps by the sequential inte-
 gration of Eqs. (7) through (10) for one time interval.  By carrying
 out this  four-step  procedure  for many  time  Intervals, one simulates the
 time history of pollutant  concentrations 1n each grid cell.
      As indicated by  Eqs.  (7) and  (8),  Steps  I and II of the solution
 procedure treat horizontal  transport due to advection and diffusion.
 Step I integrates Eq.  (7), which 1s transport of pollutants in the
 x-direction,  and Step II integrates Eq. (8),  transport 1n the y-d1rect1on.
 Since Steps  I and II  are basically  the  same except for the direction of
 integration,  only Step  I is discussed  here.   Integration of Eq. (7) is
 accomplished  by solving the advective  part  of the equation:
                    ^UHc.)  +^uAHc.) =  0     ,

 using the SHASTA method developed by Boris  and Book (1973).  This step
 is  described  1n detail  by  Reynolds  et al . (1976).  The diffusive part of
 Eq.  (7),
is then solved using a standard explicit finite difference technique.
The pollutant concentrations calculated in Step I are then Input to
Step II, which calculates the effect of transport 1n the y-d1rection.
     Equations (7) and (8) cannot be solved without specifying suitable
boundary conditions.  At any point on the horizontal boundaries, the
                                    A-8

-------
appropriate constraint 1s dependent on whether the wind  1s  flowing  Into or
out of the modeling region.  When the wind 1s  flowing  out of the  region, the
partial derivative of the pollutant concentration  with respect  to x (Step I)
or y (Step II) 1s set equal to zero.  If the wind  1s flowing Into the  region,
then the pollutant concentration must be specified along the upwind boundary.
     Step III of the numerical procedure entails the Integration  of Eq.  (9),
which considers pollutant emissions and vertical advectlon  and  diffusion.
The equation 1s solved through the use of an Implicit  finite difference
technique to eliminate stability constraints on the size of the time step
that might otherwise arise 1f the vertical grid spacing  becomes relatively
small at some point 1n a simulation.  The computation  of the vertical  wind
speed W 1s described previously by equations  (4) through (6).  If pollutants
are advected or entrained in through the top of the modeling region, the
user must specify the concentration at each species at the  top  of the  region.
Otherwise, the partial derivative of the pollutant concentration  with  respect
 o p 1s set equal to zero at the upper boundary.   If the modeling region has
been segmented Into layers, then the resulting sets of Eq.  (9)  are  coupled  at
the Interface between the layers.

     Integration of Eq. (10) to Include the contributions  of chemical
reactions 1s the final step of the numerical  procedure.   A Crank-Micol son
difference scheme 1s employed in Step IV, which yields a set of nonlinear
algebraic equations.  A solution to these equations is obtained by using a
Newton iterative procedure. For further discussion of the details of the
finite difference equations, the reader is again  referred to the reports
by Reynolds et al., (1973a, 1976), and to KWus  et al., (1977).
                                  A-9

-------
C.   THE ESTIMATION OF TURBULENT DIFFUSIVITIES
     Central to the treatment of turbulent diffusion processes in Eq. (2)
is the estimation of the horizontal and vertical diffusivity coefficients,
K^ and Ky.  Because of the empirical nature of the diffusivity concept,
these coefficients are difficult to measure and specify precisely.  With
regard to horizontal pollutant transport, advection tends to be significantly
more important than turbulent diffusion processes.  This dominating influence
is strengthened because local concentration gradients near point and line
sources cannot be represented in the Airshed Model owing to the use of a
grid with relatively coarse horizontal resolution.  Although the horizontal
diffusion terms have a small influence on predicted concentrations, they
have been retained in the model.  At present, KM is set to a constant value
       2
of 50 m /sec.

     In contrast to horizontal transport, turbulent diffusion is frequently
the dominating vertical transport process.  Thus, considerable attention has
been given to the development of algorithms for estimating the vertical dif-
fusivity, K...  The parameters used to calculate Ky are the stability class,
the ground-level wind speed, the surface roughness, the height of the wind
measurements, and the height of the grid cell.  The first step in calcu-
lating the diffusivity is to estimate the Monln-Obukhov length.  This is
done by relating the Monin-Obukhov length, L, to the surface roughness, ZQ,
and the stability function, S, using the following expression (Liu et al.,
1976):
                         (DI  - b2|S|  + b3S2)
                                   A-10

-------
where
     a, =
     bo =
0.004349,
0.003724,
0.5034,
0.2310,
0.0325.
     This formula 1s a result of the  best  fit of  observational data reported
by Colder (1972).  The stability function,  S, a numerical  representation of
                           i
the Pasqulll stability category, can  be  calculated as  follows:
                                                                  (14)
where
                  S1gn(ce)
                      1
                      0
                     -1
                                          ce>0
                                          co <0
The parameters cw and cfi are the wind speed  class  and  exposure  class  respectively,
and are defined as follows:
                  u
           cw= *
        •jr  »  0 $ "r 1 8 m/sec  ,

        4   ,  u  > 8 m/sec   ;
       .        i "*
           ce =
                                           daytime  insolation
        3   »  strong
        2   ,  moderate
        1   ,  slight
        0   ,  heavy overcast     day or night ,
                  -1  ,  •>£ cloud cover
                  -2     <|- cloud  cover
                                            nighttime cloudiness
 If S = 0, the Monin-Obukhov length 1s set to 1.0 x 10  meters.
                               A-11

-------
     Next, the friction velocity, u*, is calculated using the following
equation (Liu et al., 1976 ):
where up denotes the wind speed measured at a reference height, zf,  and where
                                      (stable or neutral)
(15)
     9 =

F an
* an
1 - * (— )
(J * L '
t * 
-------
to generate a theoretical concentration field downwind of a point source.
Optimal control  theory techniques have been employed to estimate the dif-
fuslvlty profiles  that cause the predicted concentration fields of Eq.  (2)
to agree most closely with the theoretical fields.  The reader 1s referred
to the reports by  Lamb (1976), Lamb, Chen, and Se1nf1eld (1975), and
Lamb et al. (1977) for further details.
     Mathematically,  the algorithms used to calculate Ky can be expressed
as follows:
                              «
     >  Stable conditions
ku*z expf -
i
u*
fz\
«*)
                                        /z\	      •            (18)
                                 1 + 4.7(f
     >  Neutral conditions
              Kv= T~ (ao + V * V2 + V3 + V4)
                       for 0 < z <.0.45[-F

              Ky = 0.01 m2/sec    for z > 0.45 (^-)    .            (20)

     >   Unstable conditions
              K  = kl 7  I O  + Q r ^ O
              Ny   "**•!  po    1

                       + 64(3c4 - 8?2 + 1) 1
                                A-13

-------
where
            c s ZT- -
 and
            f = Coriolis parameter,
            k = von Karman constant,
           z. = inversion height,
           v  = geostrophic wind component.
The coefficients a., and ^ have the following values:
                     = 7.396 x 10~4    ,     BQ =  0.152
                     = 6.082 x 10"2    .     BT  =  0.080
                     = 2.532    ,            B2 =  -0.039
                     = -1.272 x 10    ,      83 =  0.032
                     = 1.517 x 10    ,       B  =  0.020
      As noted earlier, the Airshed Model  has the capability of including an
 upper layer of grid cells  (frequently corresponding  to  the  inversion  layer)
 within the modeling region.   To  estimate  values  for  Ky  in this layer, vertical
 temperature gradient inputs  are  checked to  establish the appropriate  stability
 class of the layer.   The corresponding equation  cited above and the actual
 height of the grid  cell is then  used  to calculate

                                   A-14

-------
D.   THE TREATMENT OF ATMOSPHERIC CHEMISTRY
     In the earlier discussion of Eq.  (2),  1t was  Indicated  that chemical
effects are Incorporated Into the model  through  the  term R^, the net  rate
of production of pollutant 1  by chemical  reactions.   The net rate  of  reaction
can be calculated using a chemical  kinetic  mechanism, which  1s  a set  of  reac-
tions and rate constants that describe the  pertinent atmospheric chemical
phenomena.  The main chemistry mechanism employed  1n the Airshed Model 1s
known as the Carbon-Bond Mechanism (CBM), as reported by Whltten and  Hogo
(1977) and Whltten, Hogo, and KUlus (1979).  A  more recent  version of the
Carbon-Bond Mechanism (CBM II) 1s described 1n Whltten,  Klllus, and Hogo
(1980) and Whltten, et al. (1979).  The 65  reactions and rate constants  in
this mechanism are presented  1n Table A-l.

     A unique feature of the  Carbon-Bond Mechanism 1s Its treatment of
organic species.  Since every reaction of all the  organic species  found  1n
an urban atmosphere cannot be considered, these  pollutants must be grouped
to Hm1t the number of reactions and species to  a  manageable level while
permitting reasonable accuracy 1n predicting ozone formation.  In  the
Carbon-Bond Mechanism, each carbon atom of  an organic molecule 1s  classified
according to Its bond type.  For example, the carbon bonds in a butene
molecule (C^Hg) can be represented as follows:
                            H     H
                            I     I
                          H-C-C-C-C-H
                            l I  I l
                            H H H H
This molecule has two carbon atoms associated solely with single bonds and
two atoms associated with a double bond.  The two carbon atoms associated
with a double bond are also each associated with a single bond.   In assigning

                                A-l 5

-------
TABLE A-1.  THE CARBON-BOND MECHANISM (CBM-II)
       (WMtten, Kill us,  and Hogo, 1980)

1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.

12.
13.
14.
15.
16.
17.
18.
19.

20.

21.
22.
23.
Reaction
N02 + hv* NO + 0
0 + 02 + M-.03 + M
03 + NO •*> N02 + 02
°3 + N02 * N03 + °2
0 + N02 •* NO + 02
03 + OH -^ H02 + 02
03 + H02 -•• OH + 202
N02 + OH * HN03
09
CO + OH + H02 + C02
NO + NO + 02 •*• 2N02
N03 + NO - 2N02

N03 + N02 + H20 + 2HN03
H02 + NO -»• N02 + OH
H02 +H02 -
PAR + 0 -»• ME02 + OH
PAR + OH + ME02
OLE * 0 •*• ME02 + AC03 + X
OLE + 0 •*• CARB
OLE + OH + RA02

OLE + 03 •»• CARB + CRI6

OLE + 03 * CARB + MCR6
ETH + 0 * ME02 + H02 + CO
ETH + 0 •* CARB
Rate Constant
at 298°K
(ppm win )
Experimental
,§
2.1 x 10'*
23.9
4.8 x 10"2
1.34 x 104
7.7 x 101
5.0
1.4 x 104
2
4.4 x 10^
7.1 x 10"10
2.8 x 104
.***
311 x k (N20g + H20)
1.2x104
1.5 x 104
2 x 101
1.5 x 103
2.7 xlO3
2.7 x 103
4.2 x 104
o
8 x NT*
*
8 x 10 *
6 x 102
6 x 102
Activation
energy
(K)
—
—
1,450
2,450
—
1,000
1,525
—
—
—
—

-10.600
—
--
—
—
—
—
—

— —

—
—
—
                       A-16

-------
                       TABLE A-1.   (Continued)
Reaction
                                                  Rate Constant         Activation
                                                    at 298°K ,           energy
                                                  (ppm   nrln  )
                                                         **
                                                       /
                                                     2
24.  ETH 4- OH •*• RB02                              1.2 x  104
25.  ETH + 03 * CARB +  CRIG                      2.4 x  10"3
26.  AC03 4- NO •»• N02 +  ME02 4- C02                3.8 x  103
27.  RB02 4- NO -» N02 +  2  CARB 4- H02              1.2 x  104
28.  RA02 + NO * N02 +  2  CARB + H02              1.2 x  104
29.  ME02 + NO •*• N02 +  CARB + MEOg + X        (1.2 x 104)(A-1)/A*
30.  MEO + NO + N02 4- CARB + H02              (1.2 x
31.  ME02 + NO •*• Nitrate                          5 x 10'
32.  RB02 + 03 -»• 2 CARB + H02                    5.0
33.  RA02 + 03 * 2 CARB + H02                    2 x 102
34.  ME02 + 03 -»• CARB + H02                      5.0
35.  CARB •*• OH •»• o(H02  +  CO) +
     (1 - o) (AC03 4- X)                          (2.4  - a)  x 104
36.  CARB 4- hv -»• CO                               okf*tt
37.  CARB 4- hv -»• (1 4- 0)H02 +                      a +  1   *tf
     (1 - o)(ME02 4-X) 4- CO                          Z     f
38.  X 4- PAR -v                                    1 X  105
39.  AC03 4- N02 * PAN                             2 x  103
40.  PAN * AC03 4- N02                             2.8  x 10"2t           12,500
41.  AC03 + H02 *                                 4 x  103
42,  ME02 4- H02 -»•                                 4 x  103
43.  CRIG + NO •»• N02 4-  CARB                      1.2  x 104
44.  CRIG 4- N02 - N03 4- CARB                     8 x  103
45.  CRIG 4- CARB * Ozonlde                       2 x  103
46.  MCRG 4- NO •»• N02 +  CARB                       1.2  x 104
                                        A-17

-------
                         TABLE A-l.   (Continued)
Reaction
47.
48.
49.
50.
51.
52.
53.
54.

55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
MCRG + N02 -* N03 + CARB
MCRG + CARB -» Ozonide
CRIG H- CO
CRIG -»• Stable Products
CRIG -» 2H02 + C02
MCRG -»• Stable Products
MCRG ->• ME02 + OH + CO
MCRG -»• MEO« + HO, + CO,
Z Z Z
MCRG -»• CARB + 2H02 + CO
ARO + OH -»• ARPI + ARPI + ARPI + HOg
ARO + OH + H02 + GLY + X
ARO + OH -»- OH + GLY + W
W + CARB •>
ARPI + NO -»• NO + CARB + PAR
ARPI + NO -»• N02 + AEROSOL
ARPI + N03 -> CARB + CARB
ARPI + 03 •»• AEROSOL
GLY + OH •*• H02 + ARPI + ARPI + ARPI + CO
GLY -» ME02 + H02 + ARPI + ARPI + ARPI
Rate
at
(ppm
8 x
2 x
6.7
2.4
9 x
1.5
3.4
4.25

8.5
6 x
1.6
1.5
1.0
30
15
3.5
0.6
104
*V*I \J
vlLY
Constant
298° K
-1 mm'1)
3
3
x 102t
xl02t
10lf
x!02t
x 102t
x!02t

x 10U
103
x 103
x 104
x 105


x 104


.11
Activation
energy
(K)
—
--
--
—
—
—
«
_ _

—
--
—
—
—
—
—
—
—
—
-—
 The rate constants shown are as used to model eleven experiments at UCR that
 used mixes of seven hydrocarbons.  For that study the default values,  a =  0.5
 and A = 1.3, were used.


fUn1ts of mln'1.
§
 Un1ts of
                                   A-18

-------
                       TABLE A-1.  (Concluded)
**
  A « A 1s the average number of R0«-type radicals generated from a hydrocarbon
  between attack by OH* and generation of H0«.

fta 1s the fraction of total aldehydes that represent formaldehyde and ketones.
  kr Is the carbonyl photolysis rate constant.
                         C     1     1
      n  j. u n^ B 5 x 10   ppm" m1n~  for UCR simulations.
              '
     2 5
                                   A-19

-------
carbon atoms to bond categories 1n the CBM, precedence 1s given to double
bonds and carbonyl bonds.  The kinetic mechanism given 1n Table A-l considers
five possible types of carbon bonds:  single-bonded carbon atoms (PAR), very
reactive double bonds (OLE), aromatic rings (ARO), carbonyl bonds (CARB), and
moderately reactive double bonds  (ETH).  Single-bonded carbon atoms comprise
not only paraffin molecules, but  also portions of olefln, aromatic, aldehyde
and other molecules.  Double bonds are treated as pairs of carbon atoms.  An
activated aromatic ring  1s treated as a unit of six carbon atoms.  The carbon
atom in the carbon-oxygen double  bond of an aldehyde or ketone 1s Included
1n the carbonyl group.   To Illustrate this treatment of organic species,
1 mole of butene 1s considered to consist of 2 moles of single-bonded carbon
atoms and 1 mole of very reactive double bonds.  Table A-2 shows how other
organic species are treated 1n the Carbon Bond Mechanism.
     CBM II exhibits several Important characteristics which make 1t suitable
for incorporation 1n the Airshed  Model.  First, 1t has no adjustable parameters.
In the formulation of CBM II, special care was taken to avoid the use of
adjustable parameters that might  have a profound effect on the predicted results.
Other proposed mechanisms that do contain such parameters can sometimes be
"tuned" to predict well  for one set of conditions but must be "retuned" to fit
other conditions.  This  limits their overall utility.

     Second, CBM II was  developed to be a condensation of larger, more
detailed mechanisms for  the reactions of propylene, butane, ethylene, toluene,
formaldehyde, and acetaldehyde.   Tests Indicate that CBM II generates pre-
dictions that closely agree with  those obtained from the more detailed
mechanisms.  Moreover, this correspondence was obtained without any parameter
adjustments.  Thus, CBM  II 1s a condensation of present knowledge of the
                                   A-20

-------
                              TABLE A-2.


SPECIES             CHEMICAL RAHE
  RO.


   1        METHANE
   2        ETHANE
   3        ETHYLENE
   4        PROPANE
   6        PROPYLENE

   6        ACETYLENE
   7        CYCLOPROPANE
   8        PROPADIENE
   9        METHYLAGETYLENE
  10        'CYCLOPENTARE

  11        R-BUTANE
  12        BUTENE

  13        ISO-BUTANE
  14        1.3-BDTADIENE

  18        ETHYLACETYLENE
  16        N-PENTANE

  17        l-PENTENE   	
  18        2-METHYL-2-BUTENE
  19        HEXARE

  20        HEPTANE
  21        OCTANE

  22        NONANE
  23        'ISOMERS OF HEXARE
  24        'ISOMER8 OF HEPTANE

  26        R-DECARE
  26        'I80MER8 OF OCTANE

  27        'CYCLOHEXARE
  28        UROECARE
  29        olSOMERS OF RORARE
  30        'I80MER8 OF DECARE
  31         "ISOMERS OF URDECARE
  32        'R-DODECARE
  33         'ISOMERS OF DODECARE
  34         'R-TRIDECARE
  38         'I80MER8 OF TRIDECARE
  36         'R-TETRADECANE
  37         'ISOHERS OF TETRADECARE

  38         'R-PERTADECARE
  39         "1SONERS OF PERTADECARE
  40         'C-7 CYCLOPARAFFIRS

  41         'C-B CYCLOPARAFFIRS
  42         *C-9 CYCLOPARAFFIRS
  43         'TERPERES
  44         'METHYLCYCLOHEXARE
  48         'MIRERAL SPIRITS
CARBON-BOND PROFILES BY COMPOUND (Whltten, 1979)


   OLE          PAR         ARO          GARB
   e.ee
   e.ee
   e.ee
   e.ee
   i.ee
   v • OO
   v • Vv
   e.ee
   e.ee
   e.ee
   e.ee
   i.ee
   e.ee
   i.ee
   e.ee
   e.ee
   i.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   i.ee
   e.ee
   e.ee
 e.ee
 e.ee
 e.ee
 i.8o
 i.ee
 e.ee
 i.ee
 e. ee
 i.Bo
 3.ee
 4.ee
 2.ee
 4.ee
 e.ee
 i.ee
 s.ee
 3.ee
 3.ee
 e.ee
 7.ee
 a.ee
 9.ee
 e.ee
 7.ee
ie.ee
 a.ee
 4.ee
n.ee
 9.ee
ie.ee
ii.ee
i2.ee
i2.ee
i3.ee
is.ee
I4.ee
i4.ee
is.ee
i8.ee
 s.ee
  .00
  ,oe
 6.00
 6.00
 6.00
6,
7,
  OAA
 • W
e.ee
e.ee
e.ee
e.ee
e.ee
e.eo
e.eo
e.eo
e.ee
o.oe
e.eo
e.ee
e.ee
e.ee
e.ee
e.oe
e.ee
e.oo
e.eo
o.eo
o.eo
e.ee
fr.ee
e.oe
e.ee
e-.ee
e.oe
e.oe
e.oo
o.eo
e.oe
e.eo
e.oe
e.oe
o.oo
e.ee
e.ee
o.oe
e.oe
e.oe
o.ee
o.oo
o.ee
o.oo
                         e.ee
                         e.oe
                         e.ee
                         e.ee
                         e.ee
                         e.ee
                         e.ee
                         i.ee
                         e.ee
                         e.ee
                         e.eo
                         o.oe
                         e.ee
                         2.00
                         e.ee
                         e.ee
                         e.eo
                         2.00
                         e.oe
                         e.ee
                         e.ee
                         o.oe
                         e.ee
                         e.ee
                         e.ee
e.oo
e.oo
e.oe
e.oo
e.eo
  eika
 . w
0.00
e.oe
e.eo
e.oe
e.oo
o.oo
e.ee
e.oo
o.eo
2.00
e.ee
e.oo
                                       ETH
e.eo
o.oe
i.eo
e.oe
e.oe
e.ee
i.oo
i.oe
o.oo
i.ee
e.oo
e.oo
e.ee
e.ee
e.ee
e.oe
e.oo
e.ee
o.oo
o.oo
e.oe
o.oo
e.ee
e.ee
e.oe
o.oe
i.oo
e.oo
e.eo
e.ee
o.oo
e.eo
e.oo
o.oo
e.oe
e.oe
e.oo
o.eo
e.ee
  oe
  oe
  oo
o.eo
 .00
 .00
                                                 URREACTIVE
1.00
2.00
e.eo
1.80
e.ee
i.ee
e.ee
e.ee
i.se
e.ee
e.ee
o.oo
o.eo
e.eo
e.oo
e.eo
e.eo
e.ee
o.oo
e.eo
o.ee
e.eo
e.oo
e.eo
e.oe
e.eo
e.oo
o.oo
o.eo
o.oe
o.oo
o.oe
o.oo
o.oe
o.oo
o.eo
o.oo
o.oo
e.oo
o.oo
e.oo
e.oo
o.oo
e.ee
e.oo

-------
   TABLE  A-2 continued
NJ
K>
SPECIES

   NO.               CHEMICAL  NAME

  46        'CYCLOHEXAHORE
  47        *LACTOL SPIRITS
  48        'I8OHEBS OF BOTIHC
  49        'I80MER8 OF PElfTENE
  86        'ISOMERS OF PERTAHE
  81        METHYL ALCOHOL
  82        ETHYL ALCOHOL
  83        H-PROPYL ALCOHOL
  84        I80-PROPYL ALCOHOL
  88        H-BUTYL ALCOHOL
  86        ISO-BUTYL ALCOHOL
  87        BUTYL CBLL08OLVE

  SB        TEAT-BUTYL ALCOHOL
  89        METHYL CELLO80LVE
  60        CELL080LVE
  61        D1ACETOHE ALCOHOL
  62        ETHYL ETHER
  63        *CLYCOL ETHER
  64        'CLYCOL
  65        'PROPYLENE GLYCOL
  66        ETHYLENE GLYCOL
  67        TETRAHYDROFURAN
  68        ACETIC ACID
  69        METHYL ACETATE
  76        ETHYL ACETATE
  71        PROPYL ACETATE
  72        N-BUTYL ACETATE
  73        ETHYL ACRYLATE
  74        CELLOSOLVE ACETATE
  78        *ISOPROPYL ACETATE
  76        "METHYL AMYL ACETATE
  77        *ISOBOTYL ACETATE
  78        DIMETHYL FORMAMIDE
  79        "ISODUTYL  ISOBUTYRATE
  00        FORMALDEHYDE
  81        ACETALDEHYDE
  82        *BUTYRALDEHYDE
  83        ACETONE
  84        METHYL ETHYL KETONE
  83        METHYL N-BUTYL KETONE
  86        METHYL I80BUTYL KETONE
  87        ETIIYLENE OXIDE
  88        'PROPYLENE OXIDE

  89        ACETONITRILE
  90        ACRYLONITRILE
OLE

0.00
V • Vtr
9.99
e.ee
0.00
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee

i.*e
e.ee
e.oo
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
 0.00
                                                              PAR
                                                              3.ee
                                                              O* vV
                                                              2.00
                                                              3.00
                                                              5.00
                                                              i.ee
                                                              2.00
                                                              3.60
                                                              3.66
                                                              4.66
                                                              4.66
                                                              8.66
                                                             9.V6
                                                             2.66
                                                             9* 99
                                                             8.66
                                                             3.66
                                                              1.66
                                                              1.66
                                                              2.66
1.66
i.ee
2.66
e.ee
3.66
4.66
8.66
2.66
4.00
5.66
a.ee
6.00
e.ee
7.66
e.ee
i.ee
3.66
2.66
3.66
0.00
5.00
0.ee
2.00
  .66
  .66
                                                              g
                                                              1
ARO

 6.66
 6.66
 6.66
 e.ee
 e.ee
 6.66
 e.ee
 6.66
 e.ee
 e.ee
 e.ee
 6.66

 ff.ee
 6.66
 6.66
 6.66
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
CARB

 1.66
 6.66
 2.66
 2.66
 6.66
 6.66
 6.66
 6.66
 6.66
 6.66
 6.66
 1.66


 i!ee
 1.66
 1.66
 1.66
 i.ee
 1.66
 i.ee
 1.66
 i.ee
 e.ee
 e.ee
 i.ee
 i.ee
 1.66
 1.66
 2.00
 e.ee
 e.ee
 e.ee
 e.ee
   ee
   ee
   ee
   ee
   ee
   66
   66
   66
 6.66
 6.66
 6.66
 6.66
ETH

 1.66
e.ee
e.ee
e.ee
e.oe
e.ee
e.ee
e.ee
6.66
6.66
6.66
6.66

6166
6.66
6.66
6.66
e.ee
e.ee
6.66
e.ee
e.ee
i.ee
e.ee
e.ee
e.ee
e.ee
e. ee
 i.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 e.ee
 6.66
 1.66
UNREACTIVE

   6.66
   6.66
   6.66
   e.ee
   6.66
   e.ee
   6.06
   6.66
   e.ee
   e.ee
   6.66
   6.66

   6.«6
   6.66
   6.66
   6.66
   e.ee
   6.66
   6.66
   6.66
   6.66
   6.66
   e.ee
   s.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   e.ee
   3.00
   e.ee
   e.00
   o.oo
   0.00
   0.00
   o.oo
   0.00
   e.oo
   2.eo
   I.ee
   2.00
   0.66

-------
TABLE A-2  continued
SPECIES
   NO.
          CHEMICAL NAME
OLE
PAR
ARO
CARB
ETH
UNREACTIVE
  91
  92
  93
  94
  95
  96
  97
  98
  99
 100
 101
 102
 103
 104
 100
 106
 107
 108
 109
 110
 111
 112
 113
 114
 115
 116
 117
 118
 119
 120
 121

 1212
 123
 124
 120
 126
 127
 128
 130
ETHYLAMIRE                        0.00
TRIHETHYL AMIIfE                 .  0.00
METHYL CHLORIDE                   0.00
DICHLOROHETHARE                   0.00
CHLOROFORM                        o.oo
CAJ1UON TETRABROMIDE               0.00
'FREOH 11                         0.00
ETHYL CHLORIDE                    0.00
1.l-DICnLOROETHANE                0.00
1.1.l-TRICHLOROETHAHE             0.00
ETHYLERE DICHLORIDE               0.00
*FR£OR 12                         0.00
PERCHLOROETHYLENE                 0.00
METBYLERE BROMIDE                 0.00
1.1.2-TRICHLOROETBARE             0.00
'FREOR 113                        0.00
"TRIMETHYLFLUOROSILAKE            0.00
'MOROCHLORBERZEHE                 0.00
VINYL CHLORIDE                    0.00
HAPTHA                            0.00
BENZENE                           0.00
TOLUENE                           0.00
ETHYLBENZEHE                      0.00
1.3.5-TRIMETHYLBERZERE            0.00
STYRENE                           0.00
A-METHYLSTYRERE                   0.00
"ISOMERS OF XYLENE                0.00
'DIMKTHYLETHYLBERZERE             0.00
'1.2.3-TRIMETHYLDENZENE           0.00
'180MER9 OF ETHYLTOLOERE           .00
'I80MER9 OF BUTYLBERZERE           .00

'I8OMERS OF DIETBYLBERZERE
'I80HER8 OF TRIHETBYLBEHZERC
'ISOMER9 OF PROPYLBERZERE
PHENOLS
'XYLENE BASE ACIDS
CHLOROBERZERE
'1.4-DIOXARE

2-ETHOXYETHYL ACE1ATE
  TRICHLOROETHYLENE
             1.00
             3.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             2.00
             0.00
             0.00
             0.00
             0.00
             0.00
             e.oo
             8.00
             0.00
             1.00
             2.00
             3.00
             0.00
             1.00
             2.00
             4.00
             3.00
             3.00
             4.00

             4.00
             3.00
             3.00
             0.00
             2.00
             0.00
             2.00
             4.0
             2.0
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
            0.00
             .00
             .00
             .oe
             .00
             .00
             .00
             .00
             .00
             .00
             .00
             .00
             .00
            0.00
            1.00
            0.00
            0.0*0
            i.W
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00

              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              2.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
              .00
              .00
             0.00
             0.00
             0.00
             0.00
             0.00
             0.00
             1.00
             0.00
             0.00
             0.00
             0.00
             0.00
              1.00
              1.00
              0.00
              o.oe
              0.00
              0.00
              0.00

              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
                          r.w
              1.00
              0.00
               .00
               .00
               .00
               .00
               .00
              2.00
              2.00
              2.00
              2.00
              1.00
              0.00
              1.00
              2.00
              2.00
              3.00
              6.00
              0.00
              0.00
              6.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00
              0.00

              0.00
              0.00
              0.00
              6.00
              0.00
              6.00
              0.00



              l:tt

-------
 details  of several  relatively simple,  yet  representative, hydrocarbon
 systems.   As future detailed studies of other organic  compounds lead to
 a  better understanding of the smog formation  process,  CBM II may modified
 to incorporate these findings.

      It  is well  known that reaction rate constants  are a function of temp-
 erature.   Thus,  the model  treats  their effects on the  values of the non-
 photolysis reaction rate constants. In general, the atmosphere is expected
 to be more "reactive" on hot days than on  cold days.   Smog chamber experi-
 ments run at 101°F  generate about 100  percent more  ozone than similar
 experiments conducted at 61°F (Pitts et al.,  1977).  Given the value of a
 rate constant, Kr,  at some reference temperature, T ,  the Arrhenius relation-
 ship for the value  of k at some other  temperature T can be written as:
                         k  =  k   exp
where  Ea  is  the  activation energy of  the  reaction  and R  is the gas constant.
        a
It  should be noted  that the Arrhenius relationship predicts a smaller temp-
erature effect on ozone production than that actually observed in smog chamber
experiments.  Thus, caution should be exercised when applying the model to
situations where the ambient temperatures  significantly  exceed 30°C, the
nominal temperature at which most smog chamber experiments are performed.

     There are four photolysis reactions  1n the CBM, and the rate constant
for each of these reactions is a function  of the solar UV radiation Intensity.
Thus, the Airshed Model requires that the  user specify the temporal variations

                                      A-24

-------
of these rate constants.   Hourly values  of these  parameters  can  be  estimated
from actual solar radiation measurement  data  using  a  relationship such  as  that
proposed by Schere and Oemerjlan (1977).
     Appreciable quantities of aerosols  can cause light  scattering  effects
that alter the Intensity of UV radiation present  1n the  atmosphere.  As a  result,
the photolysis rate constant near the ground  can  be approximately 30 percent
smaller than Its value at the top of the modeling region when significant
amounts of aerosol are present.  Using predictions  of aerosol concentrations  In
a column of grid cells, the model estimates the rate constants at the bottom
and top of the modeling region.  The values at Intermediate  points  are esti-
mated by linear Interpolations.  This 1s described further 1n KWus, et al.
(1977).

E.  THE TREATMENT OF EMISSIONS
    The Airshed Model can accommodate emissions both at  the ground  and aloft.
The emissions from a particular source are Injected Into the appropriate
grid cell.  Because of the finite resolution  of the model, these emissions
are mixed Instantaneously throughout the cell.  The magnitude and spatial and
temporal distribution of emissions must be specified for all pollutants being
simulated In the model.  Furthermore, emissions of organic compounds must be
grouped according to the carbon-bond categories discussed previously.  This
step can be accomplished through the use of available emissions composition
measurements or the results of other studies  carried out 1n similar urban
areas.  An example of one such study for Los Angeles 1s reported by Trijonis
and Arledge (1976).  A large organic species data base  1s reported by  Bucon,
Macko, and Taback (1978), also based on Los Angeles.

                               A-25

-------
     Such data can  be  readily  used to apportion hydrocarbon emissions to
the  five carbon-bond categories.  The basic approach is to first compute
the  number of moles of each  organic  species emitted by a given source as
follows:
where Q    =  total  hydrocarbons emitted  (grams)
      X.j   =  weight percent of species i  in emissions Q
      M.J   =  molecular weight of species  i (grams)
      Q_   =  moles  of species i emitted  (gram-moles)
       m1

One then uses the  carbon-bond profiles  by species given in Table A-2 to
compute the  number of moles of each carbon-bond associated with each organic
species.   For example,  if a source emits 120 moles (Q_ ) of cyclohexane
                                                     m1
(entry 27  in Table A-2), this is equivalent to emitting 480 moles of single-
bonded carbon atoms (PAR) and 120 moles  of moderately reactive double bonds
(ETH).   Once the  number of moles of carbon-bonds has been computed for
each species, the  results can be summed  to compute the total number of
noles of each carbon-bond emitted by the source.
     The overall procedure is illustrated in Table A-3.  In this example, the
weight percents were obtained from Bucon, Macko, and Taback (1978) and are
representative of  architectural surface  coating emissions, primarily trade
paints.  The computations indicate that  for every 100 grams of hydrocarbons
emitted, there are 4.66 gram-moles of single-bonded carbon atoms (PAR), 0.121
g-am-moles of aromatic rings (ARO), 0.25 gram-moles of carbonyl bonds (CARB),
and 0.25 gram-moles  of moderately reactive double bonds (ETH).  To obtain the
      Note that the  ring structure of cycl oparaf f 1 ns 1s more photochemical ly
reactive than straight chain paraffins.  The opening of the ring structure
is treated as 1f the ring 1s a moderately reactive double bond.
                                A-26

-------
        TABLE A-3.   EXAMPLE COMPUTATION OF THE EMISSIONS OF THE FIVE CARBON-BONDS FOR AN INDIVIDUAL  SOURCE
      SPECIES
I
ro
N-Hexane                    86.2
Cyclohexane                 84.2
Isomers of Xylene          106.2
Toluene                     92.1
Ethyl Benzene              106.2
Acetone                     58.1
Methyl Ethyl Ketone         72.1
Methyl N-Butyl  Ketone      100.2
Methyl Isobutyl Ketone     100.2
Methyl Alcohol              32.0
Ethyl Alcohol               46.1
Isopropyl Alcohol           60.1
N-Butyl Alcohol             74.1
Isobutyl Alcohol            74.1
Propylene Glycol            76.0
Ethylene Glycol             62.1
N-Butyl Acetate            116.2
Isobutyl Acetate           116.2
Dimethyl Formamide          73.1
Isobutyl Isobutyrate       144.2
2-Ethoxyethyl Acetate      132.0

TOTAL
     WEIGHT PERCENTS
Weight
Percent
(*)
20.7
20.7
2.6
5.2
4.3
3.2
5.6
0.7
0.6
3.9
0.6
16.4
1.6
0.6
0.8
0.6
2.5
1.5
0.5
6.1
1.3
Compound

(moles)
0.240
0.246
0.024
0.056
0.041
0.055
0.078
0.007
0.006
0.122
0.013
0.273
0.022
0.008
0.011
0.010
0.022
0.013
0.007
0.042
0.010
                                                                         x 28
PAR
(moles)
1.440
0.984
0.048
0.056
0.082
0.110
0.234
0.035
0.030
0.122
0.026
0.819
0.088
0.032
0.022
0.010
0.110
0.078
0.294
0.040
4.660
x 14
ARO
(moles)
_
-
0.024
0.056
0.041
-
-
-
-
-
-
-
-
—
—
-
-
-
»
-
0.121
x 78
CARB
(moles)

-
-
-
-
0.055
0.078
0.007
0.006
-
_
-
-
-
o.on
0.010
0.022
-
0.042
0.020
0.251
x 30
ETH
(moles)
.
0.246
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
_
-
0.246
x 28
                                                                          65.2
9.4
7.5
6.9

-------
 corresponding  weight percent of each carbon-bond, the number of gram-moles 1s
 multiplied by  the weight  of the carbon-bond unit.  These weights are as follows
                      PAR   @ CH2      14 grams
                      OLE   @ C2H4     28 grams
                      ETH   @ C2H4     28 grams
                      ARO   @ CgHg     78 grams
                      CARB @ COH2     30 grams

 In this example,  the carbon-bond weight percents are therefore 65.2% PAR,
 9.4% ARO, 7.5% CARB, and  6.9% ETH.  Once such a profile of carbon -bond weight
 percents 1s developed,  1t can be readily applied to an emission rate Q to
 obtain the number of moles for each carbon-bond as follows:
                      PAR = (wt^0PAR) Q/14
                      OLE =  () Q/28
                      ETH =  f±fapL) Q/28
                      ARO =  ("Sflft^0) Q/78
                     CARB =  (WtiQQCARB) Q/30.
      Emissions  released at or near ground level are injected Into the lowest
 level of  grid cells.   Emissions emanating from elevated point sources are
 Injected  into the appropriate grid cell corresponding to the total effective
 plume rise, Ah  :
                        Ahp = Ahs + Ah,,, ,
where Ahe 1s the stack height and Ahm 1s the ultimate plume rise relative
        s                           *
to the top of the stack.  To estimate Ah^, the following algorithms recom-
mended by BHggs (1971) are employed:
                                 A-28

-------
     >   When s  >  0  and  u  >. 1 m/sec,

                           **   - 8.7            -                   (23)
     >  When s > 0 and u £ 1  m/sec and when Ah^ is less than
        Ah  predicted by Eq.  (23),
Ah
  00
                               - 8.6 Q1/4 s-3/8    .                <24)
     >  When s < 0 and Q >. 20 MWf
                           Ah  - 15.3 Q1/3 Ah*/3 iT1     .          (25)
                             oo               5
     >  When s <. 0 and Q < 20 MW,


                           Ah  * 20.8 Q1/3 X2/3 U'1     .            (26)
                             oo


In Eqs. (23) through (26),
                                s = I-?-1 I-57
                                X =

and

          Q = rate of heat emissions  (MW),
          u = wind speed at height Ah  (m/sec),
          g = gravitational acceleration  (m/sec  ),
          T = mean ambient temperature  (°K),
          e = potential temperature  (°K).
                                     A-29

-------
     The problem of plume penetration Into or through stable layers aloft
(temperature Inversions) 1s also Important.  Normally this 1s handled by
comparing the potential temperature of the plume at some elevation with
that of the ambient air at the same elevation (Briggs, 1971).  The plume
should continue to rise as long as Its potential temperature 1s higher.
Therefore, to determine the fate of the plume when it encounters a tempera-
ture inversion, provisions have been made in the Airshed Model to compute
the difference between the plume and ambient potential temperatures, A6,
using relationships suggested by Briggs (1971).  If Ae is predicted to be greater
than zero at the base of the Inversion, then the plume 1s allowed to rise until
the point where A6 = 0 is reached.  If the plume rises higher than the top of
the modeling region, then the pollutants carried in the plume are excluded
from subsequent Airshed Model calculations.
F.   THE TREATMENT OF SURFACE UPTAKE
     Pollutant removal processes are Incorporated Into the Airshed Model through
tne use of the deposition velocity concept, that 1s, the uptake flux of pollutant
1 at the surface, F^.,, is proportional to the ground-level concentration
C ,.  The proportionality constant 1s the effective deposition  velocity,
V.j.  Mathematically, the flux can be written as follows:

                         Fd1 • Vd1 Cg1                           (27)
The uptake of a pollutant takes place in three sequential stages:
     >  Transport to just above the surface by advectlon and
        turbulent diffusion.
     >  Transport to the surface Itself by processes that are
        Influenced by the shape of the surface.
     >  Absorption, adsorption, or chemical reaction at the surface.
                                A-30

-------
Thus, the overall  removal  rate  Is affected by both transport and chemical
processes.
     To estimate values  for V,.,, one  applies the concept of a resistance to
transport, Rt, and to surface removal, Rg^.  One then defines V^ as follows
When either resistance 1s  large,  V^.  approaches zero.

     To parameterize the transport  resistance, the results of studies
carried out by Owen and Thompson  (1963) and Chamberlain  (1966) are employed
to obtain:
                           Rt  = u^;2^'^;1  ,                    (29)

where u  1s the wind speed at  a reference height of 10 meters, and B
1s an empirically determined function having  the following value:
                           B"1 =  2.2  u;1/3   .                     (30)

     The surface resistance RS^,  1s dependent on the pollutant and on
the type of surface.  To determine  the surface resistance for grid square j,
    , one calculates the average  surface uptake velocity V. =  (
V
                          sU •     VkVsi     .                 (3D
                                 A-31

-------
where
     o.k = fraction of grid square j represented by land use kt
      B^ = a factor that adjusts the reference surface uptake
           velocity to that for land use k,
     ^s1 = re^erence surface uptake velocity for pollutant 1.
Values of ek for central business district, suburban residential, and rural/
agricultural land use categories have been estimated at 0.2, 0.5, and 1.0,
respectively (Killus et a!., 1977).  To satisfy the input requirements
of the Airshed Model, the user must estimate values of

                           I ajk8k
for each ground-level grid cell 1n the modeling region.  Further information
pertaining to the parameterization of pollutant removal processes, is found
in Killus et al., (1977).
                                A-32

-------
                          APPENDIX  REFERENCES


Boris, J. P., and D.  L.  Book (1973),  "Flux Corrected Transport. I. SHASTA,
    an Algorithm that Works," J.  Comp.  Phys., Vol.  II, pp. 38-69.

Brlggs, G. A. (1971), "Plume Rise:  A Recent Critical Review," Nuclear
    Safety. Vol. 12,  pp.  15-24.

Bucon, H. W., J. F. Macko, and H. J.  Taback (1978), "Volatile Organic
    Compound (VOC) Species Data Manual," EPA-450/3-78-119, KVB Engineering, Inc.
    Tustln, California.

Buslnger, J. A., and  S.  P. S. Arya  (1974), "Height of the Mixed Layer 1n the
    Stably Stratified Planetary Boundary Layer," Advances In Geophysics.
    Vol. 18A, F. N. Frenklel  and  R. E.  Munn, eds. (Academic Press, New
    York, New York).

Chamberlain, A.  C. (1966), "Transport of Gases to and from Grass and
    Grass-Like Surfaces," Proc. Roy.  Soc., Vol. A290, pp. 236-265.

Deardorff, J. W. (1972),  "Numerical Investigation of Neutral and Unstable
    Planetary Boundary Layers," J.  Atmos. Sci.. Vol. 32, pp. 1794-1807.

Golder, D. (1972), "Relations Among Stability Parameters 1n the Surface
    Layer," Bound. Layer Meteorol.. Vol. 3, pp. 47-58.

KUlus, J. P., et al., (1977), "Continued Research  in Mesoscale A1r Pollution
    Simulation Model1ng--Vol. V:  Refinements 1n Numerical Analysis, Transport,
    Chemistry, and Pollutant Removal,"  draft report for Environmental Sciences
    Research Laboratory,  Office of  Research and Development, U. S. Environmental
    Protection Agency, Contract No. 68-02-2216, Systems Applications, Incor-
    porated, San Rafael,  California.

Lamb, R. G. (1976), "Continued Research in Mesoscale A1r Pollution Simu-
    lation Modeling—Volume III:  Modeling of Mlcroscale Phenomena,"
    EPA-600/4-76-016c, Systems Applications, Incorporated, San Rafael, California,

Lamb, R. G., W.  H. Chen,  and J. H.  Seinfeld (1975), "Numer1co-Emp1r1cal Analysis
    of Atmospheric Diffusion Theories," J. Atmos. Sci.. Vol. 32, pp. 1794-1807.

Lamb, R. G., et  al.,  (1977), "Continued Research 1n Mesoscale Air Pollution
    Simulation Modeling—Vol. VI:   Further Studies  in the Modeling of Micro-
    scale Phenomena," draft report  for  Environmental Sciences Research Labora-
    tory, Office of Research and  Development, U. S. Environmental Protection
    Agency, Contract  No.  68-02-2216,  Systems Applications,  Incorporated,  San
    Rafael, California.

                                 A-33

-------
 L1u,  M.  K., et al., (1976),  "The Chemistry,  Dispersion, and Transport of Air
     Pollutants Emitted from  Fossil  Fuel  Power  Plants 1n California:  Data
     Analysis and Emission  Impact Model," Systems Applications, Incorporated,
     San  Rafael, California.

 Owen, P. R., and W. R. Thompson  (1963),  "Heat  Transfer Across Rough Surfaces,"
     J.  Fluid Mech.. Vol. 15, pp. 321-334.

 Pitts,  J. N., et al.,  (1977), "Mechanisms of Photochemical Reactions 1n
     Urban A1r—Volume  III:   Chamber Studies,"  EPA-600/3-77-014b, Environ-
     mental  Protection  Agency, Research Triangle Park, North Carolina.

 Reynolds, S. D., T. W. Tesche, and  L. E. Reid  (1978), "An Introduction to
     the SAI Airshed Model  and Its Usage," draft report to Office of Air
     Quality Planning and Standards, U. S. Environmental Protection Agency,
     EF78-53R3, Systems Applications, Inc., San Rafael, California.

 Reynolds, S. D., et al., (1976), "Continued  Research in Mesoscale A1r
     Pollution Simulation Modeling:   Volume II:  Refinements 1n the Treat-
     ments of Chemistry, Meteorology, and Numerical  Integration Procedures,"
     EPA-600/4-76-016b, Systems Applications, Incorporated, San Rafael,
     California.

          (1973a), "Further Development and Validation of a Simulation
     Model  for Estimating Ground  Level Concentrations of Photochemical
     Pollutants,"  Systems Applications,  Incorporated, San Rafael, California.

          (1973b). "Mathematical  Modeling of Photochemical A1r Pollution—I.
     Formulation of the Model," Atmos. Environ.. Vol. 7, pp. 1033-1061.

 Schere,  K.  L., and K. L. Demerjian  (1977), "A Photochemical Box Model
     for  Urban Air Quality Simulations," Proc. of the Fourth Joint Conference
     on Sensing of Environmental  Pollutants. American Chemical Society.
     6-11  November 1977, New Orleans, Louislana.

 THjonls, J.  C. and K. W. Arledge (1976), "Utility of Reactivity Criteria
     in Organic Emission Control  Strategies:  Application to the Los Angeles
     Atmosphere,"  EPA-600/3-76-091,  TRW  Environmental Services, Redondo
     Beach,  California.

 Whltten, G. Z., J.  P. Killus, and H. Hogo (1980), "Modeling of Simulated
     Photochemical  Smog with Kinetic Mechanisms, Volume I:  Final Report,"
     EPA-600/3-80-028a, Systems Applications, Inc., San Rafael, California.

 Whltten, G. Z., H.  Hogo, and J.  P.  Kill us (1979), "The Carbon-Bond
     Mechanism—A  condensed Kinetic Mechanism for Photochemical Smog,"
     Submitted for publication to Environ. Scl. Techno!., Systems
     Applications,  Inc., San Rafael, California.

Whltten, G. Z., et  al., (1979),  "Modeling of Simulated Photochemical
     Smog with  Kinetic Mechanisms, Vol.  I:  Interim Report," EPA-600/3-79-001a,
     Systems Applications, Inc.,  San Rafael, California.

Whltten,  G. Z. (1979), Unpublished.  Systems Applications, Incorporated,
    San Rafael, California.

                                  A-34

-------
Uhltten, G. Z., and H. Hogo (1978), "User's Manual for a Kinetics
    Model and Ozone Isopleth Plotting Package," Systems Applications,
    Incorporated, San Rafael, California.

Whltten, 6. Z., and H. Hogo (1977), "Mathematical  Modeling of Simulated
    Photochemical Smog," EPA-600/3-77-001. Systems Applications,
    Incorporated) San Rafael, California.
                                A-35

-------
                                    TECHNICAL REPORT DATA
                             (Please read Instructions on the reverse before completing)
 1. REPORT NO.
 EPA 450/4-80-020
                                                            3. RECIPIENT'S ACCESSION-NO.
 4. TITLE AND SUBTITLE

 Guideline  for Applying the  Airshed Model to  Urban Areas
                                          ''RtfcfobeArTE1980
                                                            6. PERFORMING ORGANIZATION CODE
 '. AUTHOR(S)
                                                            8. PERFORMING ORGANIZATION REPORT NO.
 ^PERFORMING ORGANIZATION NAME AND ADDRESS
 ronitoHng and Data Analysis Division
 Office of A1r Quality Planning and Standards
 Research  Triangle Park,  North  Carolina  27711
                                          10. PROGRAM ELEMENT NO.
                                          11. CONTRACT/GRANT NO.
 1 
-------